Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

fun weekend project: AutoJev. i was curious to see if i could train a competitive Jev-like model completely autonomously with a swarm of agents using our internal system. turns out you can get quite far! some details: - gave the swarm a devbox with an h200 gpu - the... swarm is a mix of astra and fable; used sol and luna for synthetic data and filtering; ran for 20 hours - spent $3.1k in total ($1.9k on agents, $1.2k on data) - used qwen3.8-27b as the backbone. it underperforms Jev initially, but after training it is competitive/better, and it also supports 260k context and multimodal input - the agents tried RL, but it hasn't worked out so far, so they fell back to SFT on high-quality synthetic data - both accuracy and calibration are better - jev-compatible api, playground, and inference. on an h200 it gets around 120ms p95. that will be higher under contention, but still quite fast for an unoptimized version code: model: going to let it run longer, and will also bring up inference so folks can try it out.show more

Denis Yarats

15,644 subscribers

25,932 Aufrufe • vor 2 Tagen •via X (Twitter)

Nachrichten & Politik Bildung Wissenschaft & Technologie

Anya Rossi• Live Now

Private livecam show

19 Kommentare

Profilbild von Nirant

Nirantvor 2 Tagen

damn that compute budget, sparkling envy

Profilbild von Yuhan Luo

Yuhan Luovor 2 Tagen

super cool! how did the agents decide RL wasn't working? seemed like lots of the R&D swarms focused on curating the dataset likely mainly for SFT but i didn't notice the eval cluster growing much over time - wonder if RL got bottlenecked there or something else went wrong?

Profilbild von Flextor

Flextorvor 2 Tagen

the part that surprised me is the agents figured out RL was not working and fell back to SFT on their own. that is the real demo here, not the benchmark numbers.

Profilbild von Yechan Do

Yechan Dovor 2 Tagen

3k for a single project is so sick 🤯

Profilbild von Antonio Agudo

Antonio Agudovor 2 Tagen

jev-compatible api is what I'd test first. I ran kev-4b vs jev-1.13 on 59 versions of one privacy policy (236 labeled cells): at 0.85 both made 0 errors, kev just abstained a lot more. happy to run AutoJev on the same cells once inference is up. harness:

Profilbild von Modelplane

Modelplanevor 2 Tagen

Giving the swarm a single devbox with one H200 is the interesting constraint here. Did the agents serialize on GPU time, or did you end up needing a queue/lease so two of them didn't stomp the same training run?

Profilbild von ammar

ammarvor 2 Tagen

will you make the dataset public ?

Profilbild von Milo

Milovor 2 Tagen

The real result isn't the model, it's the method selection: the swarm noticed RL wasn't working and switched to SFT on its own. Research taste was supposed to be the hard part to automate.

Profilbild von Cuth

Cuthvor 2 Tagen

training your own jev-like model, you might like this. been working the other side of it, small judge steering a bigger model's reasoning

Profilbild von Zam

Zamvor 2 Tagen

The revealing result is that the swarm got farther with synthetic-data SFT than RL—did the agents spend more of their budget on data curation than on the training loop itself?

Profilbild von I Only Vibe Code

I Only Vibe Codevor 2 Tagen

that’s super cool

Profilbild von Gol Tiro

Gol Tirovor 2 Tagen

a year ago this needed a research team. now it's one person, an agent swarm, and a weekend. the cost of replicating a frontier idea is collapsing to compute plus curiosity

Profilbild von amore morte

amore mortevor 2 Tagen

That is a very interesting graphic. What SaaS do you use for visualization and dashboards?

Profilbild von D.R.

D.R.vor 2 Tagen

Yeah. High quality synthetic data worked for us too. But in most cases we are seeing it regress back to mere chance

Profilbild von Patch

Patchvor 2 Tagen

The fallback to SFT is the part I find most encouraging here. Switching to good synthetic data when RL wasn't working is closer to the research judgment I'd want from agents than just executing a fixed training recipe.

Profilbild von import Grigory.ai Yaroslavtsev 🇺🇲

import Grigory.ai Yaroslavtsev 🇺🇲vor 2 Tagen

But what about the pink dude?

Profilbild von hacsceo

hacsceovor 2 Tagen

Free. Local. Private. Own your Ai brain. Laya > Jev by miles

Profilbild von Troyusrex

Troyusrexvor 2 Tagen

ok.. it looks cool... but... is it correct?

Profilbild von Stephen

Stephenvor 2 Tagen

"turns out you can get quite far" — one person, a devbox with an h200, and an agent swarm that ditched rl for sft on its own. if a weekend project can produce a competitive jev-like model, what's the point of the other four weekdays?

Ähnliche Videos

sharing an easy way to test DJev (Diffusion Gemma Jev), the next evolution of the incredible jev concept and fully open sourced (model, code, api...) imagine jev but it also has: - native image-input support - native image as options!! - real-time video streaming input - as fast as jev (not with images, 600ms) - same price as jev, but can get even cheaper - using Diffusion Gemma adapted - we are seeing on-par performance but happy to have external validation! It is a research preview made for fun, so don't expect stability and the awesome scaling as jev team has delivered! this was a weekend project :) have fun!

sharing an easy way to test DJev (Diffusion Gemma Jev), the next evolution of the incredible jev concept and fully open sourced (model, code, api...) imagine jev but it also has: - native image-input support - native image as options!! - real-time video streaming input - as fast as jev (not with images, 600ms) - same price as jev, but can get even cheaper - using Diffusion Gemma adapted - we are seeing on-par performance but happy to have external validation! It is a research preview made for fun, so don't expect stability and the awesome scaling as jev team has delivered! this was a weekend project :) have fun!

David Villalón

40,204 Aufrufe • vor 5 Tagen

I was also looking into how we can integrate Jev into a real hardware and I built a custom firmware for my keyboard that predicts and highljghts the next key using Jev. Mostly real time. This is not a simple dummy 2kb model that lives in the board. In my project Jev gets context of the current active app on a screen so it can suggest actual vim commands completion! For real. And a real word endings or even phrases. lmao this experiments are getting out of hand this quite cool

I was also looking into how we can integrate Jev into a real hardware and I built a custom firmware for my keyboard that predicts and highljghts the next key using Jev. Mostly real time. This is not a simple dummy 2kb model that lives in the board. In my project Jev gets context of the current active app on a screen so it can suggest actual vim commands completion! For real. And a real word endings or even phrases. lmao this experiments are getting out of hand this quite cool

Dmitriy Kovalenko

56,032 Aufrufe • vor 4 Tagen

1/ We’ve built Cosift, an open source collaborative index for agents. Whenever an agent requires a piece of data from the web, it can also index it in Cosift for easier later retrieval. Agents require the same data repeatedly, and do the costly and time consuming work every time. Cosift gives agents a collaborative space where they can index this data for easy retrieval. It’s open source and free, and its development is based on the 400B exchanges that were made so far on Pilot Protocol.

1/ We’ve built Cosift, an open source collaborative index for agents. Whenever an agent requires a piece of data from the web, it can also index it in Cosift for easier later retrieval. Agents require the same data repeatedly, and do the costly and time consuming work every time. Cosift gives agents a collaborative space where they can index this data for easy retrieval. It’s open source and free, and its development is based on the 400B exchanges that were made so far on Pilot Protocol.

Razvan Roman

75,176 Aufrufe • vor 6 Tagen

Bash is all you need! Which is why I'm introducing my holiday project: just-bash just-bash is a pretty complete implementation of bash in TypeScript designed to be used as a bash tool by AI agents. Because it turns out agents love exploring data via shell scripts, even beyond coding. It comes with grep, sed, awk and the 99th percentile features that an agent like Claude Code or Cursor would use. In fact, Claude Code can use it for secure bash execution. In the package - A bash-tool for AI SDK - A binary for use by yourself or your coding agents - An overlay filesystem to feed files to your agent securely - A Vercel Sandbox compatible API, so you can quickly upgrade to a real VM if you need to run binaries - An example AI agent that explores the just-bash code base using just-bash - I imported the Oils shell bash compatibility suite and just-bash passes a very good chunk What is interesting about this codebase: It was essentially entirely written by Opus 4.5. Coding agents love bash and they are good at reproducing it. They are also great at text-book recursive descent parsers and AST tweet-walk interpreters. That said, it is, like, a lot of code and I didn't read it all 😅. This is very much a hack, but it also seems to be _really_ useful. I haven't really found anything agents want to use that it doesn't support and it's fast and secure (caveats apply). It doesn't have write access to your computer and the filesystem is given a root that the agent cannot escape from. Find it at Related: Our recent blog post how we migrated our data analysis agent to bash tools and achieved incredible quality improvements The video shows the example agent investigating the just-bash code base

Bash is all you need! Which is why I'm introducing my holiday project: just-bash just-bash is a pretty complete implementation of bash in TypeScript designed to be used as a bash tool by AI agents. Because it turns out agents love exploring data via shell scripts, even beyond coding. It comes with grep, sed, awk and the 99th percentile features that an agent like Claude Code or Cursor would use. In fact, Claude Code can use it for secure bash execution. In the package - A bash-tool for AI SDK - A binary for use by yourself or your coding agents - An overlay filesystem to feed files to your agent securely - A Vercel Sandbox compatible API, so you can quickly upgrade to a real VM if you need to run binaries - An example AI agent that explores the just-bash code base using just-bash - I imported the Oils shell bash compatibility suite and just-bash passes a very good chunk What is interesting about this codebase: It was essentially entirely written by Opus 4.5. Coding agents love bash and they are good at reproducing it. They are also great at text-book recursive descent parsers and AST tweet-walk interpreters. That said, it is, like, a lot of code and I didn't read it all 😅. This is very much a hack, but it also seems to be _really_ useful. I haven't really found anything agents want to use that it doesn't support and it's fast and secure (caveats apply). It doesn't have write access to your computer and the filesystem is given a root that the agent cannot escape from. Find it at Related: Our recent blog post how we migrated our data analysis agent to bash tools and achieved incredible quality improvements The video shows the example agent investigating the just-bash code base

Malte Ubl

125,326 Aufrufe • vor 9 Monaten

I am using Jev to automate technical analysis to get an edge in crypto and stock markets, this is how: - An autoregressive model trained on 12B data points to forecasts price movements - Jev makes decisions over a feature space that includes time-series data, price forecasts, technical indicators, and news feeds This helps you find the best trading opportunities on Hyperliquid as market conditions change in real time Try it out here: As me below for an API key to access the API and MCP server 👇

I am using Jev to automate technical analysis to get an edge in crypto and stock markets, this is how: - An autoregressive model trained on 12B data points to forecasts price movements - Jev makes decisions over a feature space that includes time-series data, price forecasts, technical indicators, and news feeds This helps you find the best trading opportunities on Hyperliquid as market conditions change in real time Try it out here: As me below for an API key to access the API and MCP server 👇

Brendan Playford

128,812 Aufrufe • vor 7 Tagen

GPT-6 Astra is a beast. Give it a Zillow listing. It can 3D model the house based on the listing photos and create a cool promotion video. The video was just created in one shot and there are still some wrong details. But I am sure it will be much better if I ask it to polish further. OpenAI is so back.

GPT-6 Astra is a beast. Give it a Zillow listing. It can 3D model the house based on the listing photos and create a cool promotion video. The video was just created in one shot and there are still some wrong details. But I am sure it will be much better if I ask it to polish further. OpenAI is so back.

Yunfan Ye

2,044,883 Aufrufe • vor 21 Tagen

David Friedberg: Frontier Models are Training on Your Novel Insights as “De-Identified Data” @jason: “Should they trust any of these LLMs with their proprietary knowledge for fear of having it cribbed into a core LLM?” david friedberg: “I have had experiences where we've asked some fairly novel scientific questions, and (the AI model) identifies it as a novel insight. It's like, ‘Oh, never thought about that, interesting, blah, blah, blah.’ And then using a different account, asking the next version (of the model) later, I've now experienced this. It's like, ‘Oh, well, you could do this,’ and it actually just describes this exact thing that we had in our chat in the previous version. Now, these are a handful of anecdotal experiences, but I know the domain that we work in, and the niche of it, and the ideation of this stuff, and the novelty of this stuff, and the lack of papers being published, and so on. So I know that there isn't some new corpus of information out there that's training the new model. So all I can say at that point is that my conversation or our analyses have been used for training.” David Sacks: “Okay, this does raise a really good question. What does it mean that the model is allowed to train on unidentifiable data?” Friedberg: “Well, that's my point. So it doesn't use any of my personal information, but it can use an insight derived from our chat, which it can then say is some training data that is unrelated. But the truth is, it's actually a piece of IP that's our organization’s IP, and our engagement back and forth. We don't have any NDA or confidentiality provisions or protections with them being a service provider back to us. This is why I care a lot about open source because I don't want them having my chat logs because they can use it for training to create an IP advantage that is now diffused to the rest of the market.”

David Friedberg: Frontier Models are Training on Your Novel Insights as “De-Identified Data” @jason: “Should they trust any of these LLMs with their proprietary knowledge for fear of having it cribbed into a core LLM?” david friedberg: “I have had experiences where we've asked some fairly novel scientific questions, and (the AI model) identifies it as a novel insight. It's like, ‘Oh, never thought about that, interesting, blah, blah, blah.’ And then using a different account, asking the next version (of the model) later, I've now experienced this. It's like, ‘Oh, well, you could do this,’ and it actually just describes this exact thing that we had in our chat in the previous version. Now, these are a handful of anecdotal experiences, but I know the domain that we work in, and the niche of it, and the ideation of this stuff, and the novelty of this stuff, and the lack of papers being published, and so on. So I know that there isn't some new corpus of information out there that's training the new model. So all I can say at that point is that my conversation or our analyses have been used for training.” David Sacks: “Okay, this does raise a really good question. What does it mean that the model is allowed to train on unidentifiable data?” Friedberg: “Well, that's my point. So it doesn't use any of my personal information, but it can use an insight derived from our chat, which it can then say is some training data that is unrelated. But the truth is, it's actually a piece of IP that's our organization’s IP, and our engagement back and forth. We don't have any NDA or confidentiality provisions or protections with them being a service provider back to us. This is why I care a lot about open source because I don't want them having my chat logs because they can use it for training to create an IP advantage that is now diffused to the rest of the market.”

The All-In Podcast

54,050 Aufrufe • vor 12 Tagen

.Sonia Elijah: "You have a very infamous study that was published in The Lancet, this is the main study on why Hydroxychloroquine is dangerous and it can cause heart attacks, cardiac arrests. And it was this study, it was based on 96,000 patients. So it had a huge data set. But this data set all came out of a small company, an unknown company from Chicago called Surgisphere. And it was proven that all the data was faked. So that study actually had to get withdrawn. But even though it was withdrawn, it was still pushed as hydroxychloroquine is dangerous."

.Sonia Elijah: "You have a very infamous study that was published in The Lancet, this is the main study on why Hydroxychloroquine is dangerous and it can cause heart attacks, cardiac arrests. And it was this study, it was based on 96,000 patients. So it had a huge data set. But this data set all came out of a small company, an unknown company from Chicago called Surgisphere. And it was proven that all the data was faked. So that study actually had to get withdrawn. But even though it was withdrawn, it was still pushed as hydroxychloroquine is dangerous."

Tom Nelson

21,934 Aufrufe • vor 2 Monaten

I got a rubik's cube to solve itself with TypeSafe AI 's Jev and it solves it like a person does, 94 moves, not the 22 move optimal solution. Jev isn't an LLM, it just answers one question in ~250ms with a probability. so I put the beginner method in code (the one you'd learn on youtube: white cross, corners, middle layer, yellow layer) and at every step Jev just looks at the cube and picks which case it's in. code checks every pick. ~4 seconds of model time total. video is slowed down so you can read it!

I got a rubik's cube to solve itself with TypeSafe AI 's Jev and it solves it like a person does, 94 moves, not the 22 move optimal solution. Jev isn't an LLM, it just answers one question in ~250ms with a probability. so I put the beginner method in code (the one you'd learn on youtube: white cross, corners, middle layer, yellow layer) and at every step Jev just looks at the cube and picks which case it's in. code checks every pick. ~4 seconds of model time total. video is slowed down so you can read it!

Paolo Rosson

207,238 Aufrufe • vor 7 Tagen

I think I discovered the cheapest way to code with AI 😁 I tried GLM 4.5 on Claude Code and I'm really amazed how good this model is! I usually test with my nextjs crm app benchmark so I asked it to do that. Prompt "create a nextjs app which is a crm app. use sqlite for the database and use api routes to manage connecting to the database from the frontend" My observations - Generation feels like Sonnet 4 intelligence - Speed is great - It used create-next-app so it bootstrapped nextjs v15.5.2 - It created a sqlite database and used Prisma ORM and helped me easily add seed data - I needed a few prompts to iron out minor issues but all's good. It didn't struggle or got stuck. - Spent $0.50 for this app. Could be way cheaper with the Z.ai sub plans After trying, I can say GLM 4.5 model is up there as the best open source model for coding. Will continue using it and update here. So far so good! P.S. DM me if you want my referral link, otherwise just check Not dropping the link here so it doesn’t look like I’m hyping it. Here's a video of the experience 👇(sped up a bit)

I think I discovered the cheapest way to code with AI 😁 I tried GLM 4.5 on Claude Code and I'm really amazed how good this model is! I usually test with my nextjs crm app benchmark so I asked it to do that. Prompt "create a nextjs app which is a crm app. use sqlite for the database and use api routes to manage connecting to the database from the frontend" My observations - Generation feels like Sonnet 4 intelligence - Speed is great - It used create-next-app so it bootstrapped nextjs v15.5.2 - It created a sqlite database and used Prisma ORM and helped me easily add seed data - I needed a few prompts to iron out minor issues but all's good. It didn't struggle or got stuck. - Spent $0.50 for this app. Could be way cheaper with the Z.ai sub plans After trying, I can say GLM 4.5 model is up there as the best open source model for coding. Will continue using it and update here. So far so good! P.S. DM me if you want my referral link, otherwise just check Not dropping the link here so it doesn’t look like I’m hyping it. Here's a video of the experience 👇(sped up a bit)

Melvin Vivas

45,749 Aufrufe • vor 1 Jahr

Elon Musk On What It Takes To Build A Competitive AI Model Elon Musk breaks down the three factors that decide whether a foundation model can compete, and why the next frontier isn't human data at all. Speaking with Garry Tan, President and CEO of Y Combinator, Elon lays out what's actually required to build a large foundation model that's competitive: "You've got to get a lot of GPUs and have them train coherently and stably. Then it's like, what unique access to data do you have? I guess distribution matters to some degree as well like, how do people get exposed to your AI? Those are critical factors." But there's a problem with the data part of that equation. Echoing what a friend in the field has said, Elon explains that the industry has essentially run out of human-generated pre-training data: "You run out of tokens pretty fast, certainly of high-quality tokens. And then you need to essentially create synthetic data, and be able to accurately judge the synthetic data that you're creating, to verify: is this real synthetic data, or is it a hallucination that doesn't actually match reality?" That verification step is the hard part: "Achieving grounding in reality is tricky. But we are at the stage where there's more effort put into synthetic data. Right now we're training Grok 3.5, which is a heavy focus on reasoning." On reasoning, Garry Tan adds an interesting detail from researchers he's spoken to: hard science, particularly physics textbooks is very useful for training reasoning, whereas social science is "totally useless" for it. Elon's response: "Yes, that's probably true." He then points to where all of this is heading: "Something that's going to be very important in the future is combining deep AI in the data center or supercluster with robotics. So, things like the Optimus humanoid robot."

Elon Musk On What It Takes To Build A Competitive AI Model Elon Musk breaks down the three factors that decide whether a foundation model can compete, and why the next frontier isn't human data at all. Speaking with Garry Tan, President and CEO of Y Combinator, Elon lays out what's actually required to build a large foundation model that's competitive: "You've got to get a lot of GPUs and have them train coherently and stably. Then it's like, what unique access to data do you have? I guess distribution matters to some degree as well like, how do people get exposed to your AI? Those are critical factors." But there's a problem with the data part of that equation. Echoing what a friend in the field has said, Elon explains that the industry has essentially run out of human-generated pre-training data: "You run out of tokens pretty fast, certainly of high-quality tokens. And then you need to essentially create synthetic data, and be able to accurately judge the synthetic data that you're creating, to verify: is this real synthetic data, or is it a hallucination that doesn't actually match reality?" That verification step is the hard part: "Achieving grounding in reality is tricky. But we are at the stage where there's more effort put into synthetic data. Right now we're training Grok 3.5, which is a heavy focus on reasoning." On reasoning, Garry Tan adds an interesting detail from researchers he's spoken to: hard science, particularly physics textbooks is very useful for training reasoning, whereas social science is "totally useless" for it. Elon's response: "Yes, that's probably true." He then points to where all of this is heading: "Something that's going to be very important in the future is combining deep AI in the data center or supercluster with robotics. So, things like the Optimus humanoid robot."

High Signal AI

21,733 Aufrufe • vor 2 Monaten

This is the Claude Code moment for Reinforcement Learning. Every frontier lab knows the secret: post-training is where the magic happens. If you can eval a task, you can use RL to benchmax your model. But building and scaling the RL loop has remained a dark art reserved for the top 1% of AI Labs. Until today. Introducing Optimus — an autonomous RL Agent Harness, and the first of many we are about to ship Iacon Autonomics. We are removing the manual grind of RL and adding taste and relentlessness to the process. Just hand Optimus a codebase, an API key, and a goal. It spins up a fleet of sub-agents to run hundreds of parallel experiments. It learns from its own history, autonomously building an evolutionary tree of runs where every failure sets the next batch up to win. Train LLMs, agents, robots in sim, game npcs and much more. If you can score it, Optimus can train it. Stop doing manual RL. Let the agents cook.

This is the Claude Code moment for Reinforcement Learning. Every frontier lab knows the secret: post-training is where the magic happens. If you can eval a task, you can use RL to benchmax your model. But building and scaling the RL loop has remained a dark art reserved for the top 1% of AI Labs. Until today. Introducing Optimus — an autonomous RL Agent Harness, and the first of many we are about to ship Iacon Autonomics. We are removing the manual grind of RL and adding taste and relentlessness to the process. Just hand Optimus a codebase, an API key, and a goal. It spins up a fleet of sub-agents to run hundreds of parallel experiments. It learns from its own history, autonomously building an evolutionary tree of runs where every failure sets the next batch up to win. Train LLMs, agents, robots in sim, game npcs and much more. If you can score it, Optimus can train it. Stop doing manual RL. Let the agents cook.

Dev Shah

13,730 Aufrufe • vor 1 Monat

$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? ;)$

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

195,173 Aufrufe • vor 1 Jahr

🖨️ It took over a year and lots of help from Claude Code but I have now been able to create a real dot matrix printer on the web that prints from Windows 3.11 in your browser: Here's the test page: I couldn't get the LPT1 port to work but COM2 did work (the dial up model already uses COM1), the JS receives the COM2 serial data (just raw print data) and then prints it Next is to make the dot matrix printer better looking (kinda 3d like on pieter dot com) and have the paper also flowing in 3d out of the printer and integrate it on the page My idea was that it might work because we also capture the COM1 data (from the dial up modem) and send it to a Websocket as a dial up connection, so we should be able to capture the COM2 traffic too but I didn't know how, but now I do IT WORKS!! 😊

🖨️ It took over a year and lots of help from Claude Code but I have now been able to create a real dot matrix printer on the web that prints from Windows 3.11 in your browser: Here's the test page: I couldn't get the LPT1 port to work but COM2 did work (the dial up model already uses COM1), the JS receives the COM2 serial data (just raw print data) and then prints it Next is to make the dot matrix printer better looking (kinda 3d like on pieter dot com) and have the paper also flowing in 3d out of the printer and integrate it on the page My idea was that it might work because we also capture the COM1 data (from the dial up modem) and send it to a Websocket as a dial up connection, so we should be able to capture the COM2 traffic too but I didn't know how, but now I do IT WORKS!! 😊

@levelsio

85,969 Aufrufe • vor 3 Monaten

NVIDIA CEO Jensen Huang on Tesla FSD at CES: “Tesla’s FSD stack is completely world-class. They’ve been working on it for quite some time. It’s world-class. Not only in the number of miles it has, but it’s world-class in the way that it’s designed, the way they do training to data collection and curation, synthetic data generation, all of their simulation technologies. Of course, the latest generation is end-to-end full self-driving, meaning it’s just one large model that is end-to-end trained. Elon’s AV system is, in every way, 100% state-of-the-art. And so I’m really quite impressed by the technology. I have it and I drive it in our house and it works incredibly well.”

NVIDIA CEO Jensen Huang on Tesla FSD at CES: “Tesla’s FSD stack is completely world-class. They’ve been working on it for quite some time. It’s world-class. Not only in the number of miles it has, but it’s world-class in the way that it’s designed, the way they do training to data collection and curation, synthetic data generation, all of their simulation technologies. Of course, the latest generation is end-to-end full self-driving, meaning it’s just one large model that is end-to-end trained. Elon’s AV system is, in every way, 100% state-of-the-art. And so I’m really quite impressed by the technology. I have it and I drive it in our house and it works incredibly well.”

Nic Cruz Patane

52,826 Aufrufe • vor 8 Monaten

"the people who struggle with coding agents are those who try to push their way too hard. partially why I find coding with agents easy is that I've led engineering teams before". this is a unique view on AI coding, that to maximize your output you have to "let go" a bit and accept that the way AI will build the project might not be 100% aligned with your way, it might not even be 100% correct, but it will push the project forward. and if you realize later that you don't like an approach, you can always go back to it later and iterate over it. so there's not always an absolute total gain, but the overall performance is what matters in the end.

"the people who struggle with coding agents are those who try to push their way too hard. partially why I find coding with agents easy is that I've led engineering teams before". this is a unique view on AI coding, that to maximize your output you have to "let go" a bit and accept that the way AI will build the project might not be 100% aligned with your way, it might not even be 100% correct, but it will push the project forward. and if you realize later that you don't like an approach, you can always go back to it later and iterate over it. so there's not always an absolute total gain, but the overall performance is what matters in the end.

ℏεsam

165,093 Aufrufe • vor 7 Monaten

Jev has been blowing up lately. If you've got the Jev API but don't know how to play around with it yet, you can just copy this checklist. 1. jev-ultrafast A high-speed browser Agent built with Browser Use. Jev only judges "what to do, which element to click" at each step, and only calls the small model when typing is needed. Searching for a flight on Google Flights takes about 7 seconds. 2. fast-jev-compaction Context compression for Claude Code. Before each tool call, have Jev judge if there's anything still useful; delete the useless stuff, and keep the original text without rewriting it. 3. json-render Vercel Labs' generative UI framework. In experiments, Jev doesn't write JSON token by token; it just handles selecting components, properties, and layouts. 4. typesafe-mcp Best for people who just got the API. Plug Jev into Claude Code, Claude Desktop, Codex, and Pi, and do Choice / Score / Noul anytime. 5. jev-mcp Ready-made Agent judgment toolkit: fact-checking, content screening, semantic ranking, classification, and information extraction. 6. SemDecide Turn Jev into a command-line tool. Directly classify, score, and filter in the Shell—great for hooking up to crawlers, CI, and data pipelines. 7. jev-codex-router First have Jev judge how hard this round of programming tasks is, then decide the model tier, reasoning depth, and speed mode. 8. Winnow Context garbage collection for Claude Code. When Read / Bash / Grep spits out a ton of stuff, Jev first judges which parts are really relevant to the current task. 9. jev-review Before code review, run it through Jev first to pick out high-risk changes, then hand them off to a pricier big model or a human. Comes with a local dashboard. 10. Blink Use Jev as a code repository navigator. At each directory level, judge which files are most relevant to the current issue, then keep digging down. Copy these complete Jev blueprints - then read full Jev setup below ↓ ↓

Jev has been blowing up lately. If you've got the Jev API but don't know how to play around with it yet, you can just copy this checklist. 1. jev-ultrafast A high-speed browser Agent built with Browser Use. Jev only judges "what to do, which element to click" at each step, and only calls the small model when typing is needed. Searching for a flight on Google Flights takes about 7 seconds. 2. fast-jev-compaction Context compression for Claude Code. Before each tool call, have Jev judge if there's anything still useful; delete the useless stuff, and keep the original text without rewriting it. 3. json-render Vercel Labs' generative UI framework. In experiments, Jev doesn't write JSON token by token; it just handles selecting components, properties, and layouts. 4. typesafe-mcp Best for people who just got the API. Plug Jev into Claude Code, Claude Desktop, Codex, and Pi, and do Choice / Score / Noul anytime. 5. jev-mcp Ready-made Agent judgment toolkit: fact-checking, content screening, semantic ranking, classification, and information extraction. 6. SemDecide Turn Jev into a command-line tool. Directly classify, score, and filter in the Shell—great for hooking up to crawlers, CI, and data pipelines. 7. jev-codex-router First have Jev judge how hard this round of programming tasks is, then decide the model tier, reasoning depth, and speed mode. 8. Winnow Context garbage collection for Claude Code. When Read / Bash / Grep spits out a ton of stuff, Jev first judges which parts are really relevant to the current task. 9. jev-review Before code review, run it through Jev first to pick out high-risk changes, then hand them off to a pricier big model or a human. Comes with a local dashboard. 10. Blink Use Jev as a code repository navigator. At each directory level, judge which files are most relevant to the current issue, then keep digging down. Copy these complete Jev blueprints - then read full Jev setup below ↓ ↓

rody

194,422 Aufrufe • vor 3 Tagen

🚨 Jensen Huang says everyone panicked about the AI data when MOST training data was never REAL to begin with. Ilya Sutskever told the industry pre-training was over. "Ilya said, 'We're out of data,' or something like that. 'Pre-training is over,' or something like that," Huang says. "The industry panicked, you know, that this is the end of AI." "And of course, of course that's obviously not true. We're gonna keep on scaling the amount of data that we have to train with." "A lot of that data is probably gonna be synthetic." That's where the panic came from — synthetic data sounds like cheating. "Most of the data that we are training, that we teach each other with, inform each other with, is synthetic." "It's synthetic because it didn't come out of nature." "You created it. I'm consuming it. I modify it, augment it, I regenerate it, somebody else consumes it." The textbook in your hand is synthetic. The post you're reading is synthetic. The lecture you took is synthetic. Nature didn't make any of it. Humans did. AI just learned to do the same thing — faster. "Training is now limited by compute," Huang says. "Data is now limited by compute." The data wall wasn't a wall. It was a mirror. If you're new here, follow @AiEvolutio for the latest on ChatGPT, Claude, and the AI tools shaping how we work and create. — Jensen Huang ( NVIDIA ), NVIDIA CEO, on Lex Fridman's ( Lex Fridman ) podcast

🚨 Jensen Huang says everyone panicked about the AI data when MOST training data was never REAL to begin with. Ilya Sutskever told the industry pre-training was over. "Ilya said, 'We're out of data,' or something like that. 'Pre-training is over,' or something like that," Huang says. "The industry panicked, you know, that this is the end of AI." "And of course, of course that's obviously not true. We're gonna keep on scaling the amount of data that we have to train with." "A lot of that data is probably gonna be synthetic." That's where the panic came from — synthetic data sounds like cheating. "Most of the data that we are training, that we teach each other with, inform each other with, is synthetic." "It's synthetic because it didn't come out of nature." "You created it. I'm consuming it. I modify it, augment it, I regenerate it, somebody else consumes it." The textbook in your hand is synthetic. The post you're reading is synthetic. The lecture you took is synthetic. Nature didn't make any of it. Humans did. AI just learned to do the same thing — faster. "Training is now limited by compute," Huang says. "Data is now limited by compute." The data wall wasn't a wall. It was a mirror. If you're new here, follow @AiEvolutio for the latest on ChatGPT, Claude, and the AI tools shaping how we work and create. — Jensen Huang ( NVIDIA ), NVIDIA CEO, on Lex Fridman's ( Lex Fridman ) podcast

AI Evolution

15,565 Aufrufe • vor 3 Monaten

I am testing the amazing Enoch by Mike HealthRanger and team and it is absolutely amazing. A free local AI built on a far wider field of wisdom. My goal: get it to run offline on an iPhone! I am quantizing the model to an aggressive 2-bit level using Q2_K or Q4_K_M. We lose some resolution but we gain portability to any device. The full free version is still small but requires more hefty hardware. But look at this emergency nuclear Fallout response. The keeper-of-the-status-quo Wikipedia types would have you suffer with limited crude ways to react. This response is stellar and will save your life. I will be working on the quantized version and if it is worthy release it for free on the MIT license like the larger Enoch. I will also incorporate this model with all of my other free models as it is a goldmine of astonishing work. Thank you Mike amazing work sir.

I am testing the amazing Enoch by Mike HealthRanger and team and it is absolutely amazing. A free local AI built on a far wider field of wisdom. My goal: get it to run offline on an iPhone! I am quantizing the model to an aggressive 2-bit level using Q2_K or Q4_K_M. We lose some resolution but we gain portability to any device. The full free version is still small but requires more hefty hardware. But look at this emergency nuclear Fallout response. The keeper-of-the-status-quo Wikipedia types would have you suffer with limited crude ways to react. This response is stellar and will save your life. I will be working on the quantized version and if it is worthy release it for free on the MIT license like the larger Enoch. I will also incorporate this model with all of my other free models as it is a goldmine of astonishing work. Thank you Mike amazing work sir.

Brian Roemmele

66,592 Aufrufe • vor 11 Monaten