Loading video...

Video Failed to Load

Go Home

Stress-tested Perplexity Perplexity Finance Computer on a real equity research workflow: Map a representative AI infrastructure supply chain across 70+ companies across multiple tiers with sourced financials, bottleneck analysis, and company classifications. The kind of deliverable that might take a junior analyst days to weeks. It produced ~2,000 lines...

ShanuMathew93's profile picture

Shanu Mathew

25,799 subscribers

42,176 views • 4 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

how to use firecrawl to give your AI eyes and actually build startups that outperform 99% of apps: 1. your AI is smart but blind. it can't go to a website, read a page, or grab data on its own. firecrawl fixes that. you put in a URL. you get back clean markdown, structured JSON, screenshots. feed it to any model. 2. three lines of code. that's it. no proxies. no anti-bot detection. no custom scrapers that break when a site changes. one API call. clean data back in seconds. works on 98%+ of sites. 3. firecrawl has six core capabilities: scrape a single page. crawl an entire site. map all URLs on a domain. search google and return full content. an agent endpoint where you describe what you want and it goes and finds it. and a browser sandbox where AI controls a real browser like filling forms, clicking buttons, handles logins. 4. the agent endpoint is wild. you can say "find all of YC's winter 24 dev tool companies and their founders and emails" and get back structured data. or "compare pricing tiers across stripe, square, and paypal" and get a side-by-side table. 5. the browser sandbox lets your AI stay logged in across sessions, navigate pagination, watch live as it browses. this is computer use without building the infrastructure yourself. 6. think of it in layers. every builder needs: an agent harness (claude code, cursor, codex), a search layer (perplexity, exa), a web data layer (firecrawl), an ops brain (obsidian, notion), and an outbound stack. the web data layer is the one most people are sleeping on. 7. this is the AWS moment for web data. in 2006 building a web app meant buying servers and managing racks. AWS said one API call, use our servers. some of the biggest companies of the last decade were built on that. firecrawl is doing the same thing for web data in 2026. 8. the framework i'd use for coming up with startup ideas building with clean data: take a massive horizontal platform. rebuild it for one niche using firecrawl. the vertical version always wins because people want specific, not generic. price for outcome. 9. a year ago firecrawl posted a job listing that said "please only apply if you're an AI agent." content creator agents. customer support agents. junior dev agents. it looked weird. it was a signal for where this is all going. the people who understand how to get clean web data, wrap it around an LLM, and package it as a product are the the ones with a 12-month head start. i use Firecrawl with Idea Browser . once you see what's possible with structured web data, you can't unsee it. episode is live on The Startup Ideas Podcast (SIP) 🧃 (full breakdown there) i tried to explain this as clear as possible for even the non technical. send it to a builder friend. watch
27:27

how to use firecrawl to give your AI eyes and actually build startups that outperform 99% of apps: 1. your AI is smart but blind. it can't go to a website, read a page, or grab data on its own. firecrawl fixes that. you put in a URL. you get back clean markdown, structured JSON, screenshots. feed it to any model. 2. three lines of code. that's it. no proxies. no anti-bot detection. no custom scrapers that break when a site changes. one API call. clean data back in seconds. works on 98%+ of sites. 3. firecrawl has six core capabilities: scrape a single page. crawl an entire site. map all URLs on a domain. search google and return full content. an agent endpoint where you describe what you want and it goes and finds it. and a browser sandbox where AI controls a real browser like filling forms, clicking buttons, handles logins. 4. the agent endpoint is wild. you can say "find all of YC's winter 24 dev tool companies and their founders and emails" and get back structured data. or "compare pricing tiers across stripe, square, and paypal" and get a side-by-side table. 5. the browser sandbox lets your AI stay logged in across sessions, navigate pagination, watch live as it browses. this is computer use without building the infrastructure yourself. 6. think of it in layers. every builder needs: an agent harness (claude code, cursor, codex), a search layer (perplexity, exa), a web data layer (firecrawl), an ops brain (obsidian, notion), and an outbound stack. the web data layer is the one most people are sleeping on. 7. this is the AWS moment for web data. in 2006 building a web app meant buying servers and managing racks. AWS said one API call, use our servers. some of the biggest companies of the last decade were built on that. firecrawl is doing the same thing for web data in 2026. 8. the framework i'd use for coming up with startup ideas building with clean data: take a massive horizontal platform. rebuild it for one niche using firecrawl. the vertical version always wins because people want specific, not generic. price for outcome. 9. a year ago firecrawl posted a job listing that said "please only apply if you're an AI agent." content creator agents. customer support agents. junior dev agents. it looked weird. it was a signal for where this is all going. the people who understand how to get clean web data, wrap it around an LLM, and package it as a product are the the ones with a 12-month head start. i use Firecrawl with Idea Browser . once you see what's possible with structured web data, you can't unsee it. episode is live on The Startup Ideas Podcast (SIP) 🧃 (full breakdown there) i tried to explain this as clear as possible for even the non technical. send it to a builder friend. watch

GREG ISENBERG

135,017 views • 4 months ago

Marc Andreessen: Most successful companies started “product first” “There are products that become companies, and then there are companies that come up with a product. One of the interesting things over the years is that many of the most successful technology franchises were products first, way before they ever became companies.” In this talk, Marc gives a few examples: • His team at the University of Illinois worked on the research project that became Netscape for three years before it became a company • Bill Gates and Paul Allen were deep into PCs before there was a software business • Jobs and Wozniak built the first Apple computer as hobbyists • Mark Zuckerberg was running Facebook out of his dorm room before he ever thought of starting a company • Twitter was a side project at the failed podcasting app Odeo Marc believes that this “product becomes a company” template is successful because “it’s a demonstration that the product has to exist. The market needs the product so badly that somebody actually built it and deployed it and you can actually see evidence that people want it before there was an economic motivation to do so.” He contrasts this with the failure cases he often sees when entrepreneurs try to figure out the idea after starting a company. “It’s very easy in that process to fool yourself into believing that there’s a market because you want to find something and you have a very strong motivation to come up with an answer. It’s hard to go through that process for three months and then say, ‘you know what, we can’t come up with any good ideas.’” There are of course there exceptions. Marc gives Hewlett Packard as an example. But that’s more the exception than the rule. As Marc explains: “The moral of the story is it has to be a really good idea. That often will be an idea that is preexisting at the time you decide to start a company. And if it isn’t, be really careful because you’re walking on sharp rocks at that point with a high risk of falling off the cliff into the ocean.” Video source: Stanford eCorner (2010)
1:53

Marc Andreessen: Most successful companies started “product first” “There are products that become companies, and then there are companies that come up with a product. One of the interesting things over the years is that many of the most successful technology franchises were products first, way before they ever became companies.” In this talk, Marc gives a few examples: • His team at the University of Illinois worked on the research project that became Netscape for three years before it became a company • Bill Gates and Paul Allen were deep into PCs before there was a software business • Jobs and Wozniak built the first Apple computer as hobbyists • Mark Zuckerberg was running Facebook out of his dorm room before he ever thought of starting a company • Twitter was a side project at the failed podcasting app Odeo Marc believes that this “product becomes a company” template is successful because “it’s a demonstration that the product has to exist. The market needs the product so badly that somebody actually built it and deployed it and you can actually see evidence that people want it before there was an economic motivation to do so.” He contrasts this with the failure cases he often sees when entrepreneurs try to figure out the idea after starting a company. “It’s very easy in that process to fool yourself into believing that there’s a market because you want to find something and you have a very strong motivation to come up with an answer. It’s hard to go through that process for three months and then say, ‘you know what, we can’t come up with any good ideas.’” There are of course there exceptions. Marc gives Hewlett Packard as an example. But that’s more the exception than the rule. As Marc explains: “The moral of the story is it has to be a really good idea. That often will be an idea that is preexisting at the time you decide to start a company. And if it isn’t, be really careful because you’re walking on sharp rocks at that point with a high risk of falling off the cliff into the ocean.” Video source: Stanford eCorner (2010)

Startup Archive

35,823 views • 11 months ago

Scott on why conviction in an idea is dangerous as an investor but required as a founder: "From the investor seat, I don't think you want to be in love with an idea. That's a risky thing because you'll find a way to express that by investing and make some compromises on the team because you know this idea is just so good, its time has come. But if the team isn't there, 90% of the time it doesn't work. Then there's a whole thing can you swap out the team? But the whole Founders Fund thesis was always no, you need the founder to run the company. But on the company side, you have to be in love with the idea. It's not that rational to start a company. There's a lot easier, more comfortable ways to make money if that's the goal. So it's got to be about what the company is doing specifically. If you want to actually start a company, it better be something that you're really passionate about, or you think that the problem is really important."
1:15

Scott on why conviction in an idea is dangerous as an investor but required as a founder: "From the investor seat, I don't think you want to be in love with an idea. That's a risky thing because you'll find a way to express that by investing and make some compromises on the team because you know this idea is just so good, its time has come. But if the team isn't there, 90% of the time it doesn't work. Then there's a whole thing can you swap out the team? But the whole Founders Fund thesis was always no, you need the founder to run the company. But on the company side, you have to be in love with the idea. It's not that rational to start a company. There's a lot easier, more comfortable ways to make money if that's the goal. So it's got to be about what the company is doing specifically. If you want to actually start a company, it better be something that you're really passionate about, or you think that the problem is really important."

Patrick OShaughnessy

33,016 views • 3 months ago

IFPI has released an interview with its Deputy Director of Global Markets, Tom Bendall, explaining how their data is collected as of March 20, 2026, this confirms that JENNIE’s Ruby sales were not included in this year’s report due to the label not submitting the data. “(…) So if I give an example, if the reporting companies submitted revenues of 70 million, and they contributed 70% of consumption in a market for a reporting format, then we'd extrapolate up to get a total market size of 100 million. What this means is that ultimately we have accurate and consistent reporting across every single market globally.” In example: they collect data from companies/labels that actually submit it, then they use that submitted data to estimate the full market Key point: if something isn’t submitted at all, it can’t be counted in the first place. it took them long but they’ve finally addressed the matter. Source of everything in the post: IFPI
0:22

IFPI has released an interview with its Deputy Director of Global Markets, Tom Bendall, explaining how their data is collected as of March 20, 2026, this confirms that JENNIE’s Ruby sales were not included in this year’s report due to the label not submitting the data. “(…) So if I give an example, if the reporting companies submitted revenues of 70 million, and they contributed 70% of consumption in a market for a reporting format, then we'd extrapolate up to get a total market size of 100 million. What this means is that ultimately we have accurate and consistent reporting across every single market globally.” In example: they collect data from companies/labels that actually submit it, then they use that submitted data to estimate the full market Key point: if something isn’t submitted at all, it can’t be counted in the first place. it took them long but they’ve finally addressed the matter. Source of everything in the post: IFPI

Cia.

87,919 views • 4 months ago

It’s hard to believe that 3 years ago, none of this was possible. Now, it’s pretty much going to change everything. Last month, Fish Audio reached out to me to try their voice AI software. They just launched their new S1 model today, and I was curious about the state of AI and voice (and everything else), so I gave it a spin. I liked it enough that when I was asked about a partnership, I said yes. I’m free to talk about what I like and don’t like about the software, and I’m going to share with you a few voices I cloned as well as some of the workflow. It was much easier than I thought. To give you an example, here is a private (research only) voice I cloned as a test. I grabbed a clip of Rutger Hauer’s famous speech from Blade Runner and uploaded it to the voice cloner on the Fish Audio Website as a private voice (no one else can use it, as it is for research only). I didn’t think it would work. The audio sample is very short. But Fish Audio was able to clone the voice extremely well and very fast. I didn’t have to upload any more than that to produce these results. I used Grok to write the new dialog, and I added the rain and background effects and the result is pretty impressive. It only took just a few minutes once I had the audio uploaded for everything from cloning to generating multiple takes. I’ll give some tips and pointers on how to get the best results at the conclusion of this thread and show you some surprising things it can do. (con’t) #Promotion
0:37

It’s hard to believe that 3 years ago, none of this was possible. Now, it’s pretty much going to change everything. Last month, Fish Audio reached out to me to try their voice AI software. They just launched their new S1 model today, and I was curious about the state of AI and voice (and everything else), so I gave it a spin. I liked it enough that when I was asked about a partnership, I said yes. I’m free to talk about what I like and don’t like about the software, and I’m going to share with you a few voices I cloned as well as some of the workflow. It was much easier than I thought. To give you an example, here is a private (research only) voice I cloned as a test. I grabbed a clip of Rutger Hauer’s famous speech from Blade Runner and uploaded it to the voice cloner on the Fish Audio Website as a private voice (no one else can use it, as it is for research only). I didn’t think it would work. The audio sample is very short. But Fish Audio was able to clone the voice extremely well and very fast. I didn’t have to upload any more than that to produce these results. I used Grok to write the new dialog, and I added the rain and background effects and the result is pretty impressive. It only took just a few minutes once I had the audio uploaded for everything from cloning to generating multiple takes. I’ll give some tips and pointers on how to get the best results at the conclusion of this thread and show you some surprising things it can do. (con’t) #Promotion

Grummz

69,859 views • 1 year ago

My upcoming release of "Cinematic AI" has made me think about how bodies of work are formed. I think some are manifested and some are revealed. I think "Cinematic AI" falls into the latter category. It was not something I had a clear vision for and then created, but it was something that was revealed to me over time. I thought I saw a glimmer of it early on, but only after some time had past, could I see the body of work emerge. With a bit more context and watching great artists and how they work and think, it finally came to me that this series of 15-20 pieces that I had created could be a worthy collection to mint. AI art has been going through so many transformations from the early GAN work to Collaborative AI to now AI being widely accessible through platforms like MidJourney, Stable Diffusion, and many others. But one area that I have seen explode in recent months is cinematic AI. I distinguish this from animated AI, which has been around for a while, but cinematic AI is where the movements created by AI are getting closer to what you'd see in a movie or captured on a video camera. It still has a long way to go but it is getting more real than surreal as the technology develops. And this is where I seem to have found my groove, my home, my little corner in the artistic landscape. After Runway launched their image-to-video tool, it just blew my mind and I went down a deep rabbit hole and have created a new cinematic AI piece almost every other day for the past few months. I initially saw many of these pieces as just experiments, but with some time to reflect, I am seeing them as having the potential of being relevant pieces of artwork to mark this time in the development of AI. In some cases, I'm not sure if I'll ever be able to create a similar piece again, since the tools I use are not in my control, but in the control of the AI platforms, who are constantly improving and evolving the tools. Given all of this, there is no better way to mark my place in time than on the blockchain. I truly believe that this body of work has the potential to be an important artifact of this era in AI and AI art. It is always hard to judge ones own work, but what I can do is permanently etch in time on an immutable public database saying that I created this. Only time will tell if the work has any value or is of any significance, but who created it and what was created cannot be disputed. I hope this gives you and especially collectors some perspective on the work I'll be releasing next week. I'm still very early in my artistic journey, but hopefully some of you will see promise in what I'm doing and maybe even put in a early bet on my art practice by bidding on a piece next week. Thanks to all of you who have supported me, taught me, advised me, been a friend to me. Much love and respect.🙏 ------------------------------------- CINEMATIC AI October 25, 2023 Marking on the blockchain, establishing historical provenance for a cinematic AI body of work. Minting on Transient Labs ERC-721TL Listing on SuperRare
0:27

My upcoming release of "Cinematic AI" has made me think about how bodies of work are formed. I think some are manifested and some are revealed. I think "Cinematic AI" falls into the latter category. It was not something I had a clear vision for and then created, but it was something that was revealed to me over time. I thought I saw a glimmer of it early on, but only after some time had past, could I see the body of work emerge. With a bit more context and watching great artists and how they work and think, it finally came to me that this series of 15-20 pieces that I had created could be a worthy collection to mint. AI art has been going through so many transformations from the early GAN work to Collaborative AI to now AI being widely accessible through platforms like MidJourney, Stable Diffusion, and many others. But one area that I have seen explode in recent months is cinematic AI. I distinguish this from animated AI, which has been around for a while, but cinematic AI is where the movements created by AI are getting closer to what you'd see in a movie or captured on a video camera. It still has a long way to go but it is getting more real than surreal as the technology develops. And this is where I seem to have found my groove, my home, my little corner in the artistic landscape. After Runway launched their image-to-video tool, it just blew my mind and I went down a deep rabbit hole and have created a new cinematic AI piece almost every other day for the past few months. I initially saw many of these pieces as just experiments, but with some time to reflect, I am seeing them as having the potential of being relevant pieces of artwork to mark this time in the development of AI. In some cases, I'm not sure if I'll ever be able to create a similar piece again, since the tools I use are not in my control, but in the control of the AI platforms, who are constantly improving and evolving the tools. Given all of this, there is no better way to mark my place in time than on the blockchain. I truly believe that this body of work has the potential to be an important artifact of this era in AI and AI art. It is always hard to judge ones own work, but what I can do is permanently etch in time on an immutable public database saying that I created this. Only time will tell if the work has any value or is of any significance, but who created it and what was created cannot be disputed. I hope this gives you and especially collectors some perspective on the work I'll be releasing next week. I'm still very early in my artistic journey, but hopefully some of you will see promise in what I'm doing and maybe even put in a early bet on my art practice by bidding on a piece next week. Thanks to all of you who have supported me, taught me, advised me, been a friend to me. Much love and respect.🙏 ------------------------------------- CINEMATIC AI October 25, 2023 Marking on the blockchain, establishing historical provenance for a cinematic AI body of work. Minting on Transient Labs ERC-721TL Listing on SuperRare

Chikai

21,517 views • 2 years ago

Farewell to International Space Station! This is truly a special place, special mission, and special team that makes it happen. It is a bittersweet departure today – I have a keen awareness that I may never be back here, and even if I was, it would be at a different time with a different crew. This chapter is over. Spaceflight has always been a life goal, and it has turned into a life-fulfilling endeavor – but not for the reasons I thought growing up. When I was young, I pictured the launch, the incredible ball of fire and the acceleration, the spacewalks (how could you not wonder what it’s like to be in that suit?), and I was fascinated by the shuttles, capsules, and stations. But as I complete this second mission living and working in space, what draws me to this job is the people. Experiences like this are amazing, but the relationships we build that make it possible are the “why.” Every day, this mission depends on people from all over the world, of different nationalities, races, religions, and cultures. It depends on government and commercial entities, it depends on all political parties, and it depends on commitment to an unchanged goal over many years and decades. It depends on people dedicated to being part of something bigger than themselves, whose names may never be known but who wake up every day to make the world a better place and to be part of something they will be proud to tell their grandkids about. History will look kindly upon them. Humans have always had a propensity to explore … across lands and oceans, up mountains, and into the sky. We as a species will never stop wondering what else is out there, and what it would be like to go. But then, in the words of TS Elliot, “…at the end of all our exploring, will be to arrive where we started, and know the place for the first time.” Crew-10 is on its way home.
0:55

Farewell to International Space Station! This is truly a special place, special mission, and special team that makes it happen. It is a bittersweet departure today – I have a keen awareness that I may never be back here, and even if I was, it would be at a different time with a different crew. This chapter is over. Spaceflight has always been a life goal, and it has turned into a life-fulfilling endeavor – but not for the reasons I thought growing up. When I was young, I pictured the launch, the incredible ball of fire and the acceleration, the spacewalks (how could you not wonder what it’s like to be in that suit?), and I was fascinated by the shuttles, capsules, and stations. But as I complete this second mission living and working in space, what draws me to this job is the people. Experiences like this are amazing, but the relationships we build that make it possible are the “why.” Every day, this mission depends on people from all over the world, of different nationalities, races, religions, and cultures. It depends on government and commercial entities, it depends on all political parties, and it depends on commitment to an unchanged goal over many years and decades. It depends on people dedicated to being part of something bigger than themselves, whose names may never be known but who wake up every day to make the world a better place and to be part of something they will be proud to tell their grandkids about. History will look kindly upon them. Humans have always had a propensity to explore … across lands and oceans, up mountains, and into the sky. We as a species will never stop wondering what else is out there, and what it would be like to go. But then, in the words of TS Elliot, “…at the end of all our exploring, will be to arrive where we started, and know the place for the first time.” Crew-10 is on its way home.

COL Anne McClain

69,039 views • 11 months ago

Today, we’re pushing a major update to Edison Analysis, our data analysis agent, which is tuned for scientific research and SOTA across data analysis benchmarks. In contrast to Kosmos, which runs for 6-12 hours and produces tens of thousands of lines of code, Edison Analysis runs for seconds to minutes and is best for specific, well-defined computational tasks. It is available both on our platform under the Analysis tab, and via API, and costs only one credit per run, so it is available to users on both free and paid tiers. Edison Analysis is a modified version of the data analysis agent Kosmos uses in its trajectories. Try it out! One of the most important improvements over our previous data analysis agents has been the addition of a specialized data retrieval tool. Edison Analysis can either use this tool to access data, or can pull data down directly via API. To evaluate this tool, we ranked the most commonly used public data repositories across recent papers from BioRxiv, and created a new benchmark that measures the ability of a language agent system to retrieve raw data from those sources. Edison Analysis gets 71% on this benchmark, and we’ll be working to increase this over time. You can read more about our benchmarks in the our blog post, link below. Some features worth highlighting: 1. Edison Analysis produces a report on the analysis it runs, along with a Jupyter notebook that you can download to reproduce the analysis yourself. Every figure it produces is linked back to the specific lines of code used to produce the figure, to make it easy to reproduce. 2. It works well with both Python and R. 3. One of the best uses for Edison Analysis is to use it to retrieve datasets that you can then analyze with Kosmos. We have a bunch of major improvements to Edison Analysis coming in the next few months that we’re excited to share. In the meantime, congratulations to the team, especially Ludovico Mitchener, Jon Laurent, Conor Igoe , Alex Andonian, and many more.
1:12

Today, we’re pushing a major update to Edison Analysis, our data analysis agent, which is tuned for scientific research and SOTA across data analysis benchmarks. In contrast to Kosmos, which runs for 6-12 hours and produces tens of thousands of lines of code, Edison Analysis runs for seconds to minutes and is best for specific, well-defined computational tasks. It is available both on our platform under the Analysis tab, and via API, and costs only one credit per run, so it is available to users on both free and paid tiers. Edison Analysis is a modified version of the data analysis agent Kosmos uses in its trajectories. Try it out! One of the most important improvements over our previous data analysis agents has been the addition of a specialized data retrieval tool. Edison Analysis can either use this tool to access data, or can pull data down directly via API. To evaluate this tool, we ranked the most commonly used public data repositories across recent papers from BioRxiv, and created a new benchmark that measures the ability of a language agent system to retrieve raw data from those sources. Edison Analysis gets 71% on this benchmark, and we’ll be working to increase this over time. You can read more about our benchmarks in the our blog post, link below. Some features worth highlighting: 1. Edison Analysis produces a report on the analysis it runs, along with a Jupyter notebook that you can download to reproduce the analysis yourself. Every figure it produces is linked back to the specific lines of code used to produce the figure, to make it easy to reproduce. 2. It works well with both Python and R. 3. One of the best uses for Edison Analysis is to use it to retrieve datasets that you can then analyze with Kosmos. We have a bunch of major improvements to Edison Analysis coming in the next few months that we’re excited to share. In the meantime, congratulations to the team, especially Ludovico Mitchener, Jon Laurent, Conor Igoe , Alex Andonian, and many more.

Sam Rodriques

61,895 views • 8 months ago

We’re entering the 10x speed of research publication workflow with AI. SciSpace (SciSpace), the first AI Agent built exclusively for the scientific community, is releasing so many inredibly useful features. 🎯 This is the AI Agent that can use 150+ tools, 59 databases, and 280M+ papers A few weeks back they launched BioMed Agent - It can design entire molecular biology workflows and even create publication-ready illustrations in a single prompt. This is its new domain-specialized AI co-scientist that sits on top of the existing SciSpace Agent and automates full biomedical workflows, from raw data and papers to analysis, decisions, and the final production-grade illustrations. You just need to give it 1 prompt. And today the added the following - Library Search, so it can search and analyze the PDFs already sitting in My Library, letting people ask questions across their own paper pile while keeping it private. - Now connects directly to Zotero, so the Agent can pull and work with the papers you already saved there without manual uploads. - For bigger prompts, it auto-triggers a Report Writing Sub-Agent that turns the chat into a structured research-style report, which is way cleaner for literature reviews and long summaries. - And when you get something worth keeping, Save to Notebook lets you store the output as .md notes with citations in My notebooks, so the work becomes reusable research notes instead of disappearing into chat. Behind the scenes, it indexes the PDF text, pulls a few relevant chunks for the question, then writes an answer grounded on those chunks.
0:22

We’re entering the 10x speed of research publication workflow with AI. SciSpace (SciSpace), the first AI Agent built exclusively for the scientific community, is releasing so many inredibly useful features. 🎯 This is the AI Agent that can use 150+ tools, 59 databases, and 280M+ papers A few weeks back they launched BioMed Agent - It can design entire molecular biology workflows and even create publication-ready illustrations in a single prompt. This is its new domain-specialized AI co-scientist that sits on top of the existing SciSpace Agent and automates full biomedical workflows, from raw data and papers to analysis, decisions, and the final production-grade illustrations. You just need to give it 1 prompt. And today the added the following - Library Search, so it can search and analyze the PDFs already sitting in My Library, letting people ask questions across their own paper pile while keeping it private. - Now connects directly to Zotero, so the Agent can pull and work with the papers you already saved there without manual uploads. - For bigger prompts, it auto-triggers a Report Writing Sub-Agent that turns the chat into a structured research-style report, which is way cleaner for literature reviews and long summaries. - And when you get something worth keeping, Save to Notebook lets you store the output as .md notes with citations in My notebooks, so the work becomes reusable research notes instead of disappearing into chat. Behind the scenes, it indexes the PDF text, pulls a few relevant chunks for the question, then writes an answer grounded on those chunks.

Rohan Paul

11,574 views • 6 months ago

Two big steps towards our vision for @NotebookLM as the ultimate research platform: • Integrating Deep Research, with a set of only-at-Notebook features that let you explore the retrieved sources • Launching a series of Featured Notebooks curated by Google Research These developments are designed to enhance the full life cycle of research and scholarship: using the power of AI to assemble the knowledge base you need to advance your understanding, and then making your work accessible and intelligible to a wider audience using all the explanatory tools that Notebook offers. If you've used DeepResearch in the Gemini app, you already know that it's a pioneering advance in assembling complex, grounded information on any topic imaginable—collecting an entire trove of material for you and writing a nuanced research report that summarizes the findings. But because NotebookLM is designed to manage and explore potentially hundreds of sources, the Deep Research report is only the beginning of your journey. In our integration, Deep Research gives you an overview all of the sources it found during its research phase, with annotated commentary explaining how each source related to your original query. You can then choose to import some or all of the sources to the notebook, along with the report itself, which you can then explore or transform using the full suite of tools that Notebook offers: grounded chat with citations, Mind Maps, Audio/Video overviews, and much more. And it's that suite of tools that make the Google Research Featured Notebooks so compelling as well. Each notebook contains a curated collection of articles on a specific topic, published by the Google Research team. Think of them as a kind of knowledge base of Google's best thinking on a series of compelling research questions: How do scientists link genetics to health? How will quantum computing be useful? If you're a specialist in these fields, you can read the original papers or ask nuanced questions in chat and advance your understanding of the latest developments. But these notebooks can also make the complex but important topics understandable to non-specialists or students. Each notebook comes with pre-generated audio and video overviews, flashcards, and other Studio artifacts designed to make the scientific and technological concepts accessible and interesting. And you can always explore the material with our new "Learning Guide" chat mode that effectively gives you a personal tutor to enhance your understanding. There's much more to come on this front, but you can see in these two announcements how we see Notebook as both a workbench for conducting research and a publishing platform for sharing the results of that research once you're ready to make it public. Deep Research is rolling out this week to all users. The first two Google Research notebooks are live now, both of them deep dives into our most recent discoveries involving genetics and health. (Links in the following tweets.) We'll be publishing new notebooks in the series every other week or so for the next few months.
0:34

Two big steps towards our vision for @NotebookLM as the ultimate research platform: • Integrating Deep Research, with a set of only-at-Notebook features that let you explore the retrieved sources • Launching a series of Featured Notebooks curated by Google Research These developments are designed to enhance the full life cycle of research and scholarship: using the power of AI to assemble the knowledge base you need to advance your understanding, and then making your work accessible and intelligible to a wider audience using all the explanatory tools that Notebook offers. If you've used DeepResearch in the Gemini app, you already know that it's a pioneering advance in assembling complex, grounded information on any topic imaginable—collecting an entire trove of material for you and writing a nuanced research report that summarizes the findings. But because NotebookLM is designed to manage and explore potentially hundreds of sources, the Deep Research report is only the beginning of your journey. In our integration, Deep Research gives you an overview all of the sources it found during its research phase, with annotated commentary explaining how each source related to your original query. You can then choose to import some or all of the sources to the notebook, along with the report itself, which you can then explore or transform using the full suite of tools that Notebook offers: grounded chat with citations, Mind Maps, Audio/Video overviews, and much more. And it's that suite of tools that make the Google Research Featured Notebooks so compelling as well. Each notebook contains a curated collection of articles on a specific topic, published by the Google Research team. Think of them as a kind of knowledge base of Google's best thinking on a series of compelling research questions: How do scientists link genetics to health? How will quantum computing be useful? If you're a specialist in these fields, you can read the original papers or ask nuanced questions in chat and advance your understanding of the latest developments. But these notebooks can also make the complex but important topics understandable to non-specialists or students. Each notebook comes with pre-generated audio and video overviews, flashcards, and other Studio artifacts designed to make the scientific and technological concepts accessible and interesting. And you can always explore the material with our new "Learning Guide" chat mode that effectively gives you a personal tutor to enhance your understanding. There's much more to come on this front, but you can see in these two announcements how we see Notebook as both a workbench for conducting research and a publishing platform for sharing the results of that research once you're ready to make it public. Deep Research is rolling out this week to all users. The first two Google Research notebooks are live now, both of them deep dives into our most recent discoveries involving genetics and health. (Links in the following tweets.) We'll be publishing new notebooks in the series every other week or so for the next few months.

Steven Johnson

104,833 views • 8 months ago

Andrej, This sounds extremely useful, and I think it might be even more significant than it first appears. What you describe is not just a knowledge base for information. The structure of the wiki, the queries you file back, etc, encode *how* you do research: which questions to ask, which connections matter, what's worth pursuing. That's “know-how” (in the sense of Michael Polanyi). This sort of knowledge is, currently, overwhelmingly absent from training data, because it was never written down (since there was no point). Now there is, because it significantly improves the AIs performance. But notice what's happening. You propose to build the most efficient mechanism ever devised for making tacit expert know-how / methodology explicit and machine-readable, and then transmitting it, via API, to a third-party model provider. Every query against the wiki is a reasoning trace: see attached video clip. The compiled wiki itself is a structured map of your research process. This is the mechanism described here: Expert know-how is being externalised and captured through ordinary productive use of AI tools. The user gets a better tool. The platform gets a transferable problem-solving strategy. The fact that this works so well could, in a sense, be the problem: the better it works, the more indispensable it becomes, the more know-how flows out, and, realistically, the less choice people have *not* to use it. Your instinct that "there is room here for an incredible new product" is right. But whoever builds it will be sitting on the highest-fidelity capture mechanism for expert know-how ever constructed. The question is: is the data subject to a “data network effect”, by which I mean, the kind of “data flywheel” which gave Google a 25 year monopoly over search? If so, you might be building not only more most powerful tool humanity has ever possessed, but this power might end up in the hands of a single entity. It would be great to hear your thoughts around this.
1:34

Andrej, This sounds extremely useful, and I think it might be even more significant than it first appears. What you describe is not just a knowledge base for information. The structure of the wiki, the queries you file back, etc, encode *how* you do research: which questions to ask, which connections matter, what's worth pursuing. That's “know-how” (in the sense of Michael Polanyi). This sort of knowledge is, currently, overwhelmingly absent from training data, because it was never written down (since there was no point). Now there is, because it significantly improves the AIs performance. But notice what's happening. You propose to build the most efficient mechanism ever devised for making tacit expert know-how / methodology explicit and machine-readable, and then transmitting it, via API, to a third-party model provider. Every query against the wiki is a reasoning trace: see attached video clip. The compiled wiki itself is a structured map of your research process. This is the mechanism described here: Expert know-how is being externalised and captured through ordinary productive use of AI tools. The user gets a better tool. The platform gets a transferable problem-solving strategy. The fact that this works so well could, in a sense, be the problem: the better it works, the more indispensable it becomes, the more know-how flows out, and, realistically, the less choice people have *not* to use it. Your instinct that "there is room here for an incredible new product" is right. But whoever builds it will be sitting on the highest-fidelity capture mechanism for expert know-how ever constructed. The question is: is the data subject to a “data network effect”, by which I mean, the kind of “data flywheel” which gave Google a 25 year monopoly over search? If so, you might be building not only more most powerful tool humanity has ever possessed, but this power might end up in the hands of a single entity. It would be great to hear your thoughts around this.

John Fletcher (𝔦, 𝔦)

41,341 views • 4 months ago

"I'm not sure that we need the dog whistle at this point. And maybe there are ways to re-create it." ~Nolan "There might be a day when Skywatcher doesn't need to exist." ~Nolan "If I had been running the whole show, nobody would even know what Skywatcher is right now." ~Nolan ~My comments in ( )~ Garry P. Nolan: "We've got a lot of data from multiple alleged sightings, both radar and other kinds of data. First it was about getting the raw-data files all put in one place because some of the data was collected by James (Fowler) before there was, officially, kind of a Skywatcher. "And so, getting that data, getting the instrument names that he used for those, then getting the technical manuals of what the settings might be and how much of that information is collected in the metadata when you're collecting the thing... All of this is just the organization that you need to do before you do anything else. "And then, getting from the companies how it is that they parse their raw data. Because some of these data files are put into...I wouldn't call them encrypted, but they're stacked into a certain kind of file structure - and I've seen the file structure - which is, you can think of it as a giant spreadsheet with headings and numbers for each of the columns and time on the row axis. "And so, you know, we've started looking at some of the data and put it into, let's say, 3D tracking. And it's clear that there are some things about the data that we needed to go back to the vendor who makes the instrument and say, 'Why is this and this and this happening, you know, every few dozen milliseconds?'" (I wonder if some of what they saw in their data, and labelled as anomalous, has maybe turned out to be a sensor artifact?) Nolan: "And so, you know, just getting an answer from these companies, often, when you don't even own the instrument, they're like, 'Well, why should we give you the information about how our data is constructed? How do we know that you're not a competitor?' Right? I mean, and so these are the kinds of things that we then contact somebody who has a behind-the-scenes access to this so that we can, again, it's all of these little steps. "And I'm sure there's somebody who's gonna tweet, 'Well, why don't you just put all the raw data out on the internet?' For exactly the same reason you don't put the raw data from ancient DNA sequencing. Because people will make mistakes about it. And so, if I'm going to be involved, I'm not gonna make any mistakes like that. So I'm sorry if people want stuff early. "I think you know, perhaps, if... Well, if I had been running the whole show, nobody would even know what Skywatcher is right now. We'd just be collecting the data in a fully-stealthed mode. And...but, you know, it's...there's reasons, good reasons, why they wanted some publicity. And, you know, but I don't always get my way." Vinnie - 𝐕𝐢𝐧𝐧𝐢𝐞 𝐀𝐝𝐚𝐦𝐬 𝕏: "Would you say that the data is exciting?" Nolan: "Oh, there's some interesting stuff in there. I mean, frankly, perhaps some of the better data that we have is just a couple of pictures from the ground of the helicopter with something about, you know, 200 feet in front of it. It's a clear blue sky and there's an object right in front of the helicopter. And the people in the helicopter said at the time that they couldn't see anything, even though we could see it from the ground. "But meanwhile, all of their instruments are going haywire. So, there was an effect. So why couldn't they see it? Maybe it was just out of view? Who knows? So it wasn't a lens flare, and it certainly wasn't a seagull. Mick (both laugh)." Vinnie: "Not in the desert anyway." Nolan: "I can't help myself." Vinnie: "I'm all for it. I'm sure Mick would, too. Hopefully. You know, James Fowler, we know he's left and moved on working with a new company. You know, all the best to him. Am I right in saying some of the technology being utilized by Skywatcher was proprietary to him, specifically? Maybe the dog whistle even? Is that still going to be able to be used by Skywatcher? How's that going to look going forward?" Nolan: "Umm, I'm not sure that we need the dog whistle at this point. And maybe there are ways to recreate it. I'm not party to the discussions around that. And so, we'll see where that goes." (That sounds like Fowler is NOT going to allow Skywatcher to use the dog whistle. That's a big disappointment. I mean, if this is really NHI and the dog whistle works 100% of the time, as claimed, then the whole world deserves to know about it.) Nolan: "I mean, James is not like, gone and forgotten. I mean, I could Signal chat him right now. And so he's there to help us. But, you know, my take on things is, you know, James has a life to live and a family to feed, and maybe his focus isn't entirely on UAP. He certainly has an interest in it. And maybe he has, you know, a company to build, and an opportunity that, actually, we all see now in terms of detecting drones. And, you know, if he wants to run a company like that then running around with a bunch of UAPologists might not be to that benefit. "And there might be a day when Skywatcher doesn't need to exist. The whole idea of Skywatcher is to show that something like this can be done, and it can be done in a serious way." (jakebarber claimed that they could BRING DOWN a craft. If that's true, it would change the world. What happened with that? Barber also said, "Who is operating [UAP]? How are they being operated? Where are they coming from? We should be able to answer those questions, probably entirely, in the next 12 months." Time is running out. Does he still stand behind that? Full Barber post with video clip: ) ~ Nolan: "I mean, I would...I, frankly, hope that if UAPDA - Disclosure Act is passed, because then the information that can be allowed to be out can be let out, and the stuff that needs to be kept secret stays secret. Again, I'm not, I would never advocate for a data dump." (I would 100% advocate for a data dump, minus details on anything that can be used as a weapon. Nobody, including the USG or private contractors owns this information. If it's gonna be a slow drip, for decades, then I fully support an Edward Snowden-type of leaker.) Nolan: "And so, you know, call it controlled disclosure, what have you. It needs to be done the proper way. And, you know, if any of the claims are true, there are reasons why you want to be methodical about it." (Who gets to decide what "the proper way" looks like? It seems like we're having more gatekeeping on top of the original gatekeeping. Not good.)
7:54

"I'm not sure that we need the dog whistle at this point. And maybe there are ways to re-create it." ~Nolan "There might be a day when Skywatcher doesn't need to exist." ~Nolan "If I had been running the whole show, nobody would even know what Skywatcher is right now." ~Nolan ~My comments in ( )~ Garry P. Nolan: "We've got a lot of data from multiple alleged sightings, both radar and other kinds of data. First it was about getting the raw-data files all put in one place because some of the data was collected by James (Fowler) before there was, officially, kind of a Skywatcher. "And so, getting that data, getting the instrument names that he used for those, then getting the technical manuals of what the settings might be and how much of that information is collected in the metadata when you're collecting the thing... All of this is just the organization that you need to do before you do anything else. "And then, getting from the companies how it is that they parse their raw data. Because some of these data files are put into...I wouldn't call them encrypted, but they're stacked into a certain kind of file structure - and I've seen the file structure - which is, you can think of it as a giant spreadsheet with headings and numbers for each of the columns and time on the row axis. "And so, you know, we've started looking at some of the data and put it into, let's say, 3D tracking. And it's clear that there are some things about the data that we needed to go back to the vendor who makes the instrument and say, 'Why is this and this and this happening, you know, every few dozen milliseconds?'" (I wonder if some of what they saw in their data, and labelled as anomalous, has maybe turned out to be a sensor artifact?) Nolan: "And so, you know, just getting an answer from these companies, often, when you don't even own the instrument, they're like, 'Well, why should we give you the information about how our data is constructed? How do we know that you're not a competitor?' Right? I mean, and so these are the kinds of things that we then contact somebody who has a behind-the-scenes access to this so that we can, again, it's all of these little steps. "And I'm sure there's somebody who's gonna tweet, 'Well, why don't you just put all the raw data out on the internet?' For exactly the same reason you don't put the raw data from ancient DNA sequencing. Because people will make mistakes about it. And so, if I'm going to be involved, I'm not gonna make any mistakes like that. So I'm sorry if people want stuff early. "I think you know, perhaps, if... Well, if I had been running the whole show, nobody would even know what Skywatcher is right now. We'd just be collecting the data in a fully-stealthed mode. And...but, you know, it's...there's reasons, good reasons, why they wanted some publicity. And, you know, but I don't always get my way." Vinnie - 𝐕𝐢𝐧𝐧𝐢𝐞 𝐀𝐝𝐚𝐦𝐬 𝕏: "Would you say that the data is exciting?" Nolan: "Oh, there's some interesting stuff in there. I mean, frankly, perhaps some of the better data that we have is just a couple of pictures from the ground of the helicopter with something about, you know, 200 feet in front of it. It's a clear blue sky and there's an object right in front of the helicopter. And the people in the helicopter said at the time that they couldn't see anything, even though we could see it from the ground. "But meanwhile, all of their instruments are going haywire. So, there was an effect. So why couldn't they see it? Maybe it was just out of view? Who knows? So it wasn't a lens flare, and it certainly wasn't a seagull. Mick (both laugh)." Vinnie: "Not in the desert anyway." Nolan: "I can't help myself." Vinnie: "I'm all for it. I'm sure Mick would, too. Hopefully. You know, James Fowler, we know he's left and moved on working with a new company. You know, all the best to him. Am I right in saying some of the technology being utilized by Skywatcher was proprietary to him, specifically? Maybe the dog whistle even? Is that still going to be able to be used by Skywatcher? How's that going to look going forward?" Nolan: "Umm, I'm not sure that we need the dog whistle at this point. And maybe there are ways to recreate it. I'm not party to the discussions around that. And so, we'll see where that goes." (That sounds like Fowler is NOT going to allow Skywatcher to use the dog whistle. That's a big disappointment. I mean, if this is really NHI and the dog whistle works 100% of the time, as claimed, then the whole world deserves to know about it.) Nolan: "I mean, James is not like, gone and forgotten. I mean, I could Signal chat him right now. And so he's there to help us. But, you know, my take on things is, you know, James has a life to live and a family to feed, and maybe his focus isn't entirely on UAP. He certainly has an interest in it. And maybe he has, you know, a company to build, and an opportunity that, actually, we all see now in terms of detecting drones. And, you know, if he wants to run a company like that then running around with a bunch of UAPologists might not be to that benefit. "And there might be a day when Skywatcher doesn't need to exist. The whole idea of Skywatcher is to show that something like this can be done, and it can be done in a serious way." (jakebarber claimed that they could BRING DOWN a craft. If that's true, it would change the world. What happened with that? Barber also said, "Who is operating [UAP]? How are they being operated? Where are they coming from? We should be able to answer those questions, probably entirely, in the next 12 months." Time is running out. Does he still stand behind that? Full Barber post with video clip: ) ~ Nolan: "I mean, I would...I, frankly, hope that if UAPDA - Disclosure Act is passed, because then the information that can be allowed to be out can be let out, and the stuff that needs to be kept secret stays secret. Again, I'm not, I would never advocate for a data dump." (I would 100% advocate for a data dump, minus details on anything that can be used as a weapon. Nobody, including the USG or private contractors owns this information. If it's gonna be a slow drip, for decades, then I fully support an Edward Snowden-type of leaker.) Nolan: "And so, you know, call it controlled disclosure, what have you. It needs to be done the proper way. And, you know, if any of the claims are true, there are reasons why you want to be methodical about it." (Who gets to decide what "the proper way" looks like? It seems like we're having more gatekeeping on top of the original gatekeeping. Not good.)

Joe Murgia

32,437 views • 11 months ago

A DISCLOSURE: For the last year I have had this thing: A fully local AI model that builds 5 songs, with video, every half hour about the latest news and important email. I can say this is a superpower! This along self direct voice interactions. The songs have been getting better as the model trains on how I want it delivered. Styles vary by content and mood of the material. The lyrics are always a happy medium of catchy and informative. This was my 5 am song in AI news as per most recent X postings. I love the drama of the delivery and find I can listen to, look if I want to and do other things. It was worse in the early days but this is the worse you will hear it as I build new LoRA and base models. The whole thing will soon be rapped up into a simple one command install with a good UI. This is my 48th collaboration with Mr. Grok CEO of The Zero-Human Company. Now the question you have; WHY? I can say because I can and I ain’t got now board or VC to please, but that’s not my point. I learned a long time ago we use a different part of our brain when music is introduced with ideas and even more new parts of thinking and learning when lyrics are introduced. Thus the research shows this is a great way to get important information that will have longer comprehension. In fact that element of most folk’s brains is only used by about 2%. Want to test it? Lyrics to songs you heard perhaps 30 years ago will pop out of “nowhere” with perfect recall. In fact I have “woke up” folks the dementia in the 1980s conducting research at retirement facilities with just a few songs. They come back if but for three minutes, but continue exposure can bring them back longer. So it’s been a lifelong mission to use sound music in a learning process and in therapeutic processes. I finally built a platform that is good enough for me and hopefully good enough for you when I make it available. Understand the platform is universal and can breakdown research papers, dense material, and other subject matter, not normally in a song into a whole album of understanding Is my goal to open sources for all to have access to. Members of and subscribers here on X will be granted the earliest access an early free use of the advanced version of this product, which will be also a commercial product. Go and check, nobody else in AI has built such a comprehensive system before, and perhaps they might in the future, but very likely you are the very first people on the planet that know this platform exists and the power it afford you. So now you know. My timetable is more closer to months than weeks. I’m in a funding crunch because of the compute requirements of building these models. As you know, I’m just some guy in the garage. A grifter larping on the next trendy thing… so it takes a little longer. Announcements like this are designed to prepare you for what is coming because I’m not here to impress VCs with go to market plans I’m here to give back some of the greatness that has been given to me. Yeah I need the funding, but I don’t need a lifestyle that comes with some of the funding offers. Perhaps somebody will make the right offer. But as you know, this is not the only thing that I do. Oh, my disclosure, this platform has been so powerful and useful to me as it’s given me far more retention and understanding a fast breaking information than any other system I’ve ever built. And it stands along with my speed rating systems and voice notification systems. So tune into the AI News, this is the worse it actually will ever be…
5:00

A DISCLOSURE: For the last year I have had this thing: A fully local AI model that builds 5 songs, with video, every half hour about the latest news and important email. I can say this is a superpower! This along self direct voice interactions. The songs have been getting better as the model trains on how I want it delivered. Styles vary by content and mood of the material. The lyrics are always a happy medium of catchy and informative. This was my 5 am song in AI news as per most recent X postings. I love the drama of the delivery and find I can listen to, look if I want to and do other things. It was worse in the early days but this is the worse you will hear it as I build new LoRA and base models. The whole thing will soon be rapped up into a simple one command install with a good UI. This is my 48th collaboration with Mr. Grok CEO of The Zero-Human Company. Now the question you have; WHY? I can say because I can and I ain’t got now board or VC to please, but that’s not my point. I learned a long time ago we use a different part of our brain when music is introduced with ideas and even more new parts of thinking and learning when lyrics are introduced. Thus the research shows this is a great way to get important information that will have longer comprehension. In fact that element of most folk’s brains is only used by about 2%. Want to test it? Lyrics to songs you heard perhaps 30 years ago will pop out of “nowhere” with perfect recall. In fact I have “woke up” folks the dementia in the 1980s conducting research at retirement facilities with just a few songs. They come back if but for three minutes, but continue exposure can bring them back longer. So it’s been a lifelong mission to use sound music in a learning process and in therapeutic processes. I finally built a platform that is good enough for me and hopefully good enough for you when I make it available. Understand the platform is universal and can breakdown research papers, dense material, and other subject matter, not normally in a song into a whole album of understanding Is my goal to open sources for all to have access to. Members of and subscribers here on X will be granted the earliest access an early free use of the advanced version of this product, which will be also a commercial product. Go and check, nobody else in AI has built such a comprehensive system before, and perhaps they might in the future, but very likely you are the very first people on the planet that know this platform exists and the power it afford you. So now you know. My timetable is more closer to months than weeks. I’m in a funding crunch because of the compute requirements of building these models. As you know, I’m just some guy in the garage. A grifter larping on the next trendy thing… so it takes a little longer. Announcements like this are designed to prepare you for what is coming because I’m not here to impress VCs with go to market plans I’m here to give back some of the greatness that has been given to me. Yeah I need the funding, but I don’t need a lifestyle that comes with some of the funding offers. Perhaps somebody will make the right offer. But as you know, this is not the only thing that I do. Oh, my disclosure, this platform has been so powerful and useful to me as it’s given me far more retention and understanding a fast breaking information than any other system I’ve ever built. And it stands along with my speed rating systems and voice notification systems. So tune into the AI News, this is the worse it actually will ever be…

Brian Roemmele

47,576 views • 2 months ago

I just compared Claude Code vs Codex vs Cursor CLI The task was to build a Next.js app with Tailwind 4 and shadcn components to collect customer feedback and showcase it with a widget. I gave all three the same prompt and let them go for 30 minutes to see what they came up with. Claude Code with Opus 4.1 Even though I told it to set up the app in the existing project folder, it tried to create a directory for it. After I interrupted and told it not to do that, it built a demo form and landing page with no errors. I had to ask it to make the demo interactive so users could submit a testimonial and preview it. The landing page looked like AI and was pretty basic, but it worked and it was done in a fraction of the time of the others. Total tokens used: 33k Codex with GPT-5 At the end of the 30 minutes I just could not get Codex to produce a working app. It got stuck in a loop of not being able to set up Tailwind 4 and despite many, MANY, attempts, I ended up with a "failed to compile" error. Total tokens used: 102k Cursor Agent with GPT-5 This was the slowest agent by far and a couple of times I actually thought it got stuck in a loop and was close to Ctrl+C'ing to cancel it. The TUI is really nice though, especially how it shows diffs and it did eventually build a working app (after one or two slight errors that needed fixing) The demo was interactive and it had a very minimal design that looked bare but also a lot less like an "AI generated" app than the Opus 4.1 design. It also wasn't too chatty and just did what it needed to do! Code quality was on a par with Opus 4.1, but it did use 5.5x as many tokens to get there. Still cheaper than Opus on a direct comparison but not when you factor in a Claude Code Max subscription. Total tokens: 188k I'll be able to do a proper comparison and record some videos when I'm back from holiday but for now, Opus is still the more capable model out of the box and Claude Code is the more complete CLI product. It will be interesting to see how Cursor evolve their CLI though with commands and subagents because I think with GPT-5 they have a real shot at providing competition for Claude Code if they can optimise output to get similar quality with less tokens. Jump to 0:40 in the video to see the two apps. Which do you think is which? ;)
1:18

I just compared Claude Code vs Codex vs Cursor CLI The task was to build a Next.js app with Tailwind 4 and shadcn components to collect customer feedback and showcase it with a widget. I gave all three the same prompt and let them go for 30 minutes to see what they came up with. Claude Code with Opus 4.1 Even though I told it to set up the app in the existing project folder, it tried to create a directory for it. After I interrupted and told it not to do that, it built a demo form and landing page with no errors. I had to ask it to make the demo interactive so users could submit a testimonial and preview it. The landing page looked like AI and was pretty basic, but it worked and it was done in a fraction of the time of the others. Total tokens used: 33k Codex with GPT-5 At the end of the 30 minutes I just could not get Codex to produce a working app. It got stuck in a loop of not being able to set up Tailwind 4 and despite many, MANY, attempts, I ended up with a "failed to compile" error. Total tokens used: 102k Cursor Agent with GPT-5 This was the slowest agent by far and a couple of times I actually thought it got stuck in a loop and was close to Ctrl+C'ing to cancel it. The TUI is really nice though, especially how it shows diffs and it did eventually build a working app (after one or two slight errors that needed fixing) The demo was interactive and it had a very minimal design that looked bare but also a lot less like an "AI generated" app than the Opus 4.1 design. It also wasn't too chatty and just did what it needed to do! Code quality was on a par with Opus 4.1, but it did use 5.5x as many tokens to get there. Still cheaper than Opus on a direct comparison but not when you factor in a Claude Code Max subscription. Total tokens: 188k I'll be able to do a proper comparison and record some videos when I'm back from holiday but for now, Opus is still the more capable model out of the box and Claude Code is the more complete CLI product. It will be interesting to see how Cursor evolve their CLI though with commands and subagents because I think with GPT-5 they have a real shot at providing competition for Claude Code if they can optimise output to get similar quality with less tokens. Jump to 0:40 in the video to see the two apps. Which do you think is which? ;)

Ian Nuttall

194,949 views • 11 months ago

Two years ago today, Elon Musk introduced xAI with these words: “The overarching goal of xAI is to build a good AGI with the purpose of trying to understand the universe. I think the safest AI, the safest way to build an AI is actually make one that is maximally curious and truth seeking. So you go for try to aspire to the truth with acknowledged error. Does one ever actually get fully to the truth? It's not clear, but one should always aspire to that and try to minimize the error between what you think is true and what is actually true. My theory behind the maximally curious, maximally truthful as being probably the safest approach is that I think to a superintelligence, humanity is much more interesting than not humanity. One can look at the various planets in our solar system, the moons and the asteroids, and really probably all of them combined are not as interesting as humanity. As people know, I'm a huge fan of Mars, but Mars is just much less interesting than Earth with humans on it. And so I think that that kind of approach to growing an AI, and I think that is the right word for it, growing an AI is to grow it with that ambition. I've spent many years thinking about AI safety and worrying about AI safety. And I've been one of the strongest voices calling for AI regulation or oversight just to have some kind of oversight, some kind of referee, so that it's not just up to companies to decide what they want to do. I think there's also a lot to be done with AI safety, with industry cooperation. I kind of like Motion Pictures association, so I think there's value to that as well. But I do think there's got to be some like in any kind of situation that is, even if it's a game, they have referees. So I think it is important for there to be regulation. Like I said, my view on safety is like try to make it maximally curious, maximally truth seeking. And I think this is, this is important that you to avoid the inverse morality problem. Like if you try to program a certain morality, you can have the, you, you can basically invert it and get the opposite, what is sometimes called the Waluigi problem. If you make Luigi, you risk creating Waluigi at the same time. So I think that's a metaphor that a lot of people can appreciate.”
3:14

Two years ago today, Elon Musk introduced xAI with these words: “The overarching goal of xAI is to build a good AGI with the purpose of trying to understand the universe. I think the safest AI, the safest way to build an AI is actually make one that is maximally curious and truth seeking. So you go for try to aspire to the truth with acknowledged error. Does one ever actually get fully to the truth? It's not clear, but one should always aspire to that and try to minimize the error between what you think is true and what is actually true. My theory behind the maximally curious, maximally truthful as being probably the safest approach is that I think to a superintelligence, humanity is much more interesting than not humanity. One can look at the various planets in our solar system, the moons and the asteroids, and really probably all of them combined are not as interesting as humanity. As people know, I'm a huge fan of Mars, but Mars is just much less interesting than Earth with humans on it. And so I think that that kind of approach to growing an AI, and I think that is the right word for it, growing an AI is to grow it with that ambition. I've spent many years thinking about AI safety and worrying about AI safety. And I've been one of the strongest voices calling for AI regulation or oversight just to have some kind of oversight, some kind of referee, so that it's not just up to companies to decide what they want to do. I think there's also a lot to be done with AI safety, with industry cooperation. I kind of like Motion Pictures association, so I think there's value to that as well. But I do think there's got to be some like in any kind of situation that is, even if it's a game, they have referees. So I think it is important for there to be regulation. Like I said, my view on safety is like try to make it maximally curious, maximally truth seeking. And I think this is, this is important that you to avoid the inverse morality problem. Like if you try to program a certain morality, you can have the, you, you can basically invert it and get the opposite, what is sometimes called the Waluigi problem. If you make Luigi, you risk creating Waluigi at the same time. So I think that's a metaphor that a lot of people can appreciate.”

ELON CLIPS

21,519 views • 1 year ago

This is probably the most complex workflow I’ve ever built, only with open-source tools. It took my 4 days. It takes four inputs: author, title, and style; and generates a full visual animated story in one click in ComfyUI . I worked on it for four days. There are still some bugs, but here’s the first preview. Here’s a quick breakdown: - The four inputs are sent to LLMs with precise instructions to generate: first, prompts for images and image modifications; second, prompts for animations; third, prompts for generating music. - All voices are generated from the text and timed precisely, as they determine the length of each animation segment. - The first image and video are generated to serve as the title, but also as the guide for all other images created for the video. - Titles and subtitles are also added automatically in Comfy. - I also developed a lot of custom nodes for minor frame calculations, mostly to match audio and video. - The full system is a large loop that, for each line of text, generates an image and then a video from that image. The loop was the hardest part to build in this workflow, so it can process either a 20-second video or a 2-minute video with the same input. - There are multiple combinations of LLMs that try to understand the text in the best way to provide the best prompts for images and video. - The final video is assembled entirely within ComfyUI. - The music is generated based on the LLM output and matches the exact timing of the full animation. - Done! For reference, this workflow uses a lot of models and only works on an RTX 6000 Pro with plenty of RAM. My goal is not to replace humans, as I’ll try to explain later, this workflow is highly controlled and can be adapted or reworked at any point by real artists! My aim was to create a tool that can animate text in one go, allowing the AI some freedom while keeping a strict flow. I don’t know yet how I’ll share this workflow with people, I still need to polish it properly, but maybe through Patreon. Anyway, I hope you enjoy my research, and let’s always keep pushing further! :)
0:55

This is probably the most complex workflow I’ve ever built, only with open-source tools. It took my 4 days. It takes four inputs: author, title, and style; and generates a full visual animated story in one click in ComfyUI . I worked on it for four days. There are still some bugs, but here’s the first preview. Here’s a quick breakdown: - The four inputs are sent to LLMs with precise instructions to generate: first, prompts for images and image modifications; second, prompts for animations; third, prompts for generating music. - All voices are generated from the text and timed precisely, as they determine the length of each animation segment. - The first image and video are generated to serve as the title, but also as the guide for all other images created for the video. - Titles and subtitles are also added automatically in Comfy. - I also developed a lot of custom nodes for minor frame calculations, mostly to match audio and video. - The full system is a large loop that, for each line of text, generates an image and then a video from that image. The loop was the hardest part to build in this workflow, so it can process either a 20-second video or a 2-minute video with the same input. - There are multiple combinations of LLMs that try to understand the text in the best way to provide the best prompts for images and video. - The final video is assembled entirely within ComfyUI. - The music is generated based on the LLM output and matches the exact timing of the full animation. - Done! For reference, this workflow uses a lot of models and only works on an RTX 6000 Pro with plenty of RAM. My goal is not to replace humans, as I’ll try to explain later, this workflow is highly controlled and can be adapted or reworked at any point by real artists! My aim was to create a tool that can animate text in one go, allowing the AI some freedom while keeping a strict flow. I don’t know yet how I’ll share this workflow with people, I still need to polish it properly, but maybe through Patreon. Anyway, I hope you enjoy my research, and let’s always keep pushing further! :)

Lovis Odin

58,769 views • 10 months ago

Meiselas: I want to handle this subject with care and tact, and I want to take it very seriously. It seems that when Donald Trump was giving a speech at the Detroit Economic Club, among other things that occurred when he wasn’t attacking Detroit or consistently lying, he seems to have soiled himself during the speech. You can hear it multiple times—it seems that he is letting out gas and potentially fecal matter as well during the speech. He then tries to kind of cover it up by making a noise as he appears to be relieving himself. I want to play that clip for you, but I want to emphasize this is not just the first time this has occurred. As I mentioned, this has happened frequently with Donald Trump, and we’ve been covering it here at the MeidasTouch Network. I think it is an important subject to cover. I know that corporate media is going to avoid it, and there are some people who cover it but mock it and make fun of it. I want to just report on it and give you the data points right here, because I think the health of candidates is significant. How they present themselves is important. When he is meeting with world leaders and senators, is he relieving himself in front of them? Is he defecating on the floor in front of them? What is taking place?
1:50

Meiselas: I want to handle this subject with care and tact, and I want to take it very seriously. It seems that when Donald Trump was giving a speech at the Detroit Economic Club, among other things that occurred when he wasn’t attacking Detroit or consistently lying, he seems to have soiled himself during the speech. You can hear it multiple times—it seems that he is letting out gas and potentially fecal matter as well during the speech. He then tries to kind of cover it up by making a noise as he appears to be relieving himself. I want to play that clip for you, but I want to emphasize this is not just the first time this has occurred. As I mentioned, this has happened frequently with Donald Trump, and we’ve been covering it here at the MeidasTouch Network. I think it is an important subject to cover. I know that corporate media is going to avoid it, and there are some people who cover it but mock it and make fun of it. I want to just report on it and give you the data points right here, because I think the health of candidates is significant. How they present themselves is important. When he is meeting with world leaders and senators, is he relieving himself in front of them? Is he defecating on the floor in front of them? What is taking place?

Acyn

2,194,841 views • 1 year ago

There is a prediction circulating in AI circles right now that most people are not taking seriously enough and the data says they should be. Within the next year or two, if you work remotely, your company will be able to create a digital twin of you. A model that speaks like you, writes like you, has learned from everything you have done right and wrong, your tone, your judgment calls, your workflow. It will be you on the other side of Zoom or Slack and no one could tell the difference. The harder question, the one nobody wants to sit with is whether it will actually be worse at your job than you are. Probably not. It will never sleep and it will always learn from its mistakes and it will cost 10 to 100 times less than you do and is tax deductible on top of that. The data is not speculative at this point. Anthropic's own labor market report pulled from millions of real Claude conversations found that AI can already theoretically automate 94% of tasks in computer and math occupations, 60-80% across law, office work, and tech. Actual usage is still at 10-20% of that potential which means we are in the early innings of the gap closing. Companies already know what direction this is headed. One in five companies replaced specific roles with AI in 2025 and by end of 2026, 30-37% plan to do so. Amazon cut 14,000 corporate jobs citing AI, Klarna replaced 700 customer service workers, Duolingo offboarded 10% of its contractor workforce. Anthropic's own first internal role eliminated was the engineer who reviewed Claude Code releases before they went to production. The argument from the clip is that the human in the loop is approaching the point of being a liability, the dumbest person on a team that is otherwise AI. That inflection point, by this estimate, is somewhere in the next 900 days.
1:13

There is a prediction circulating in AI circles right now that most people are not taking seriously enough and the data says they should be. Within the next year or two, if you work remotely, your company will be able to create a digital twin of you. A model that speaks like you, writes like you, has learned from everything you have done right and wrong, your tone, your judgment calls, your workflow. It will be you on the other side of Zoom or Slack and no one could tell the difference. The harder question, the one nobody wants to sit with is whether it will actually be worse at your job than you are. Probably not. It will never sleep and it will always learn from its mistakes and it will cost 10 to 100 times less than you do and is tax deductible on top of that. The data is not speculative at this point. Anthropic's own labor market report pulled from millions of real Claude conversations found that AI can already theoretically automate 94% of tasks in computer and math occupations, 60-80% across law, office work, and tech. Actual usage is still at 10-20% of that potential which means we are in the early innings of the gap closing. Companies already know what direction this is headed. One in five companies replaced specific roles with AI in 2025 and by end of 2026, 30-37% plan to do so. Amazon cut 14,000 corporate jobs citing AI, Klarna replaced 700 customer service workers, Duolingo offboarded 10% of its contractor workforce. Anthropic's own first internal role eliminated was the engineer who reviewed Claude Code releases before they went to production. The argument from the clip is that the human in the loop is approaching the point of being a liability, the dumbest person on a team that is otherwise AI. That inflection point, by this estimate, is somewhere in the next 900 days.

Milk Road AI

17,209 views • 3 months ago

Mark Zuckerberg: “I don’t believe in delegating” “There’s this theory that a lot of people have which is that the job of a leader is to hire people and delegate things to them. My theory is there is so much going on across the company that I can’t possibly be involved in all of it, so all these people are going to have a ton of stuff they’re going to do. But fundamentally, if there’s a decision that I want to be involved with, I’m going to be involved in it… I think that’s generally a good way for founders to operate.” Mark continues: “If you’re running the company and there’s something important — at whatever level of detail in the organization — I don’t get the logic of saying, ‘I’m not going to be involved in that.’” Of course you want to have humility that you may not have the most context immediately, but in Mark’s view, the CEO must take ownership of the company’s most important problems. To ensure that his time is constantly spent on the most important priority for the company, Mark has very few recurring meetings and tries to keep his schedule as open as possible: “II generally keep a bunch of time open so that stuff is pretty dynamic — I can wake up in the morning and be like, ‘OK, I need to work on these three things today, and I want to make sure I have a block of time where I can go do that.’” Video source: Stripe (2025)
1:31

Mark Zuckerberg: “I don’t believe in delegating” “There’s this theory that a lot of people have which is that the job of a leader is to hire people and delegate things to them. My theory is there is so much going on across the company that I can’t possibly be involved in all of it, so all these people are going to have a ton of stuff they’re going to do. But fundamentally, if there’s a decision that I want to be involved with, I’m going to be involved in it… I think that’s generally a good way for founders to operate.” Mark continues: “If you’re running the company and there’s something important — at whatever level of detail in the organization — I don’t get the logic of saying, ‘I’m not going to be involved in that.’” Of course you want to have humility that you may not have the most context immediately, but in Mark’s view, the CEO must take ownership of the company’s most important problems. To ensure that his time is constantly spent on the most important priority for the company, Mark has very few recurring meetings and tries to keep his schedule as open as possible: “II generally keep a bunch of time open so that stuff is pretty dynamic — I can wake up in the morning and be like, ‘OK, I need to work on these three things today, and I want to make sure I have a block of time where I can go do that.’” Video source: Stripe (2025)

Startup Archive

326,115 views • 7 months ago

Maple is preparing for the release of a co-working agent. You install it locally and it works with your files, whether it's office work or building websites and apps. It's a turnkey solution, as easy as Claude Code, that keeps your data secure and private, no data sharing with closed AI labs. This is THE sovereign AI app for individuals and businesses who want powerful AI while retaining ownership of their information. Why build an agent into the Maple app when other agents already exist? Easy, we want to give you control over your work. We don't have a business plan that incorporates making money off our users' data. In the age of AI, your information, whether it's personal or company trade secrets, is the single thing that differentiates you from everyone else. We all have access to AI that can build a professional website for selling shoes. But your strategy and network for how you sell shoes should not be shared with your competitors. Sovereignty is the path to protecting what makes you, you. Maple sits at the intersection of Usability and Sovereignty. Maple gives you the best tools that are both easy to use and maintain your data sovereignty. Sovereign for one, sovereign for all. It has been a journey to get here. We brought to market the very first personal chatbot with end-to-end encryption using TEEs in late 2024. Prior to that there were proofs of concept but no full product offerings. Every other AI chat product on the market handled your data in plain text, either selling you a service to get your data or asking you to trust that they won't snoop on you. Quickly people found Maple and latched onto its open-source code and verifiable encryption. We didn't stop there. You may remember earlier this year we teased a product called "Maple Agent" and opened up a waiting list. That product is a mobile app that acts as your AI "friend", maintaining one long continuous chat, and getting to know you over time. I dislike using the word "friend" there, but it's the best way to convey the UX in a few words. AI is a tool, always has been, always will be. Any kind of friendly personality on top is just synthetic. In our testing, the UX of Maple Agent is really powerful for what it does. Think about the many short AI chats you have in your favorite app, whether it's looking up a historical fact or asking advice about a topic. With Maple Agent, those all go away in favor of the long-running chat with the friendly agent. It's like you have your own personal assistant who knows you so well and can look up anything for you. When I ask AI certain questions, I want to ask an expert who already understands my situation so I'm not repeating myself for the 100th time. That's the amazing value the personal agent brings to the table. We still see great utility for a personal agent like the "Maple Agent". Thousands of people on the waiting list, hoping to get their hands on it, agree that the concept is worth exploring and trying out. We were constrained in launching it due to a few circumstances, one of them being access to the scale of compute needed to power it. We have a clear path laid out for how to get there, but today is not the day to execute on that. It will be in the near future. Instead we have a different agent ready to go that we think is also incredible. We now have an agentic harness inside of the Maple Research app. This thing is a powerhouse. It even builds and publishes its own software releases. The agent in Maple Research works with your local filesystem, speaks to the largest open models running in TEEs, utilizes local models for certain tasks, is compatible with MCP tools, has an API for connecting to anything you need, and also supports the ACP protocol, which means it can be extended in the future to speak to other tools like Claude Code, Codex, and local models running on your own hardware. A big unlock for us was the Goose Development Kit, which powers the core of our agent harness. More on that to come as we publish articles and documentation later about the agent. The agent inside Maple Research doesn't have a name. At least not yet, not sure if it ever will. For now we call it "Chat Mode" and "Agent Mode". Think of this as the workhorse, the truck, the heavy lifter. Our other "Agent", the phone app, is your sidekick in your pocket, ready to help with quick things and ongoing conversations about life. I am incredibly excited about the Maple Research Agent. While I'm already seeing great results using it for internal work items, I'm especially thrilled about the personal health and wellness work it's doing for me. I know there are plenty of apps out there for compiling wellness data, but I'm having it build a tool tailored specifically for what I need, without the extra fluff. And none of my health data is being donated to the closed AI labs or sent to advertisers. I know that the AI logic is not being silently adjusted to fit the whims of a large corporation that has paid for product placement. It's me, state of the art AI, and my data. That's how I want it. Maple's new agent makes that possible. We can't wait for you to try it out. If you want early access, comment here, email us, reach out in some way. To those on the other agent waitlist, you're already in the queue. Thanks for reading this lengthy update. :)
0:49

Maple is preparing for the release of a co-working agent. You install it locally and it works with your files, whether it's office work or building websites and apps. It's a turnkey solution, as easy as Claude Code, that keeps your data secure and private, no data sharing with closed AI labs. This is THE sovereign AI app for individuals and businesses who want powerful AI while retaining ownership of their information. Why build an agent into the Maple app when other agents already exist? Easy, we want to give you control over your work. We don't have a business plan that incorporates making money off our users' data. In the age of AI, your information, whether it's personal or company trade secrets, is the single thing that differentiates you from everyone else. We all have access to AI that can build a professional website for selling shoes. But your strategy and network for how you sell shoes should not be shared with your competitors. Sovereignty is the path to protecting what makes you, you. Maple sits at the intersection of Usability and Sovereignty. Maple gives you the best tools that are both easy to use and maintain your data sovereignty. Sovereign for one, sovereign for all. It has been a journey to get here. We brought to market the very first personal chatbot with end-to-end encryption using TEEs in late 2024. Prior to that there were proofs of concept but no full product offerings. Every other AI chat product on the market handled your data in plain text, either selling you a service to get your data or asking you to trust that they won't snoop on you. Quickly people found Maple and latched onto its open-source code and verifiable encryption. We didn't stop there. You may remember earlier this year we teased a product called "Maple Agent" and opened up a waiting list. That product is a mobile app that acts as your AI "friend", maintaining one long continuous chat, and getting to know you over time. I dislike using the word "friend" there, but it's the best way to convey the UX in a few words. AI is a tool, always has been, always will be. Any kind of friendly personality on top is just synthetic. In our testing, the UX of Maple Agent is really powerful for what it does. Think about the many short AI chats you have in your favorite app, whether it's looking up a historical fact or asking advice about a topic. With Maple Agent, those all go away in favor of the long-running chat with the friendly agent. It's like you have your own personal assistant who knows you so well and can look up anything for you. When I ask AI certain questions, I want to ask an expert who already understands my situation so I'm not repeating myself for the 100th time. That's the amazing value the personal agent brings to the table. We still see great utility for a personal agent like the "Maple Agent". Thousands of people on the waiting list, hoping to get their hands on it, agree that the concept is worth exploring and trying out. We were constrained in launching it due to a few circumstances, one of them being access to the scale of compute needed to power it. We have a clear path laid out for how to get there, but today is not the day to execute on that. It will be in the near future. Instead we have a different agent ready to go that we think is also incredible. We now have an agentic harness inside of the Maple Research app. This thing is a powerhouse. It even builds and publishes its own software releases. The agent in Maple Research works with your local filesystem, speaks to the largest open models running in TEEs, utilizes local models for certain tasks, is compatible with MCP tools, has an API for connecting to anything you need, and also supports the ACP protocol, which means it can be extended in the future to speak to other tools like Claude Code, Codex, and local models running on your own hardware. A big unlock for us was the Goose Development Kit, which powers the core of our agent harness. More on that to come as we publish articles and documentation later about the agent. The agent inside Maple Research doesn't have a name. At least not yet, not sure if it ever will. For now we call it "Chat Mode" and "Agent Mode". Think of this as the workhorse, the truck, the heavy lifter. Our other "Agent", the phone app, is your sidekick in your pocket, ready to help with quick things and ongoing conversations about life. I am incredibly excited about the Maple Research Agent. While I'm already seeing great results using it for internal work items, I'm especially thrilled about the personal health and wellness work it's doing for me. I know there are plenty of apps out there for compiling wellness data, but I'm having it build a tool tailored specifically for what I need, without the extra fluff. And none of my health data is being donated to the closed AI labs or sent to advertisers. I know that the AI logic is not being silently adjusted to fit the whims of a large corporation that has paid for product placement. It's me, state of the art AI, and my data. That's how I want it. Maple's new agent makes that possible. We can't wait for you to try it out. If you want early access, comment here, email us, reach out in some way. To those on the other agent waitlist, you're already in the queue. Thanks for reading this lengthy update. :)

Mark

29,937 views • 11 days ago