正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

computer use agents are slow because they think too much they make round trips to an LLM. every single time. even for a command you've said a hundred times before. that's why it always feels slow. my mac now remembers. the first time i say something new, like "search... youtube for lofi beats", a big model figures it out once, does it, and saves the recipe as a json spec: {"op":"act","do":"open","target":"Safari"} {"op":"act","do":"type","text":"{query}"} every time after that, a tiny model on my laptop recognizes the command, fills in the parameters, and replays the recipe with CUA say it once, it's slow. say it again, it's instant. the fine tuned tiny model is 350 million parameters. its only job is recognizing which of my commands i just said. and when the big model does plan, it streams actions line by line. each line runs the moment it's finished, so the first action is executing while the model is still writing the second. everything it learns lives in one plain text file on my disk. i've been playing around a lot with Cua by Francesco and peekaboo by Peter Steinberger 🦞 . curious whether a standard computer use spec makes sense, one that streaming models can drive to make computer use actually fast.show more

Milind S

3,335 subscribers

14,235 次观看 • 3 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

42 条评论

Milind S 的头像

Milind S3 个月前

this is what im playing around with

🥔🥔🥔 的头像

🥔🥔🥔3 个月前

👀

Milind S 的头像

Milind S3 个月前

:D

joshua 的头像

joshua3 个月前

smells like there’s a big if true idea here, cool stuff

Milind S 的头像

Milind S3 个月前

Finding out in realtime :D

Queue 的头像

Queue3 个月前

the latency problem is less “the model is slow” and more “the model is being asked to rediscover the same procedure.” caching the routine path feels like the right boundary: keep novel decisions in the loop, move deterministic commands out of it.

Milind S 的头像

Milind S3 个月前

It isnt quite transparent of how caching works with LLMs. This gives me more control over things. But yes I think its a similar concept that I’m implementing

Queue 的头像

Queue3 个月前

yeah, cache hits are cute until you cannot see what it actually did.

Queue 的头像

Queue3 个月前

they also dont really disclose cache costs,. atleast to my understanding

PJ 的头像

PJ3 个月前

Good stuff, hope u can crack computer use with local models working with the cua team. Challenge - can you think of how to solve the problem of apps which don't surface accessibility trees such as most enterprise/business software? Or even apps running in VM's / remote desktops. Closest thing I can think of is using a harness which utilizes a thinking model & a model like Locate Anything 3B. But you are the expert here, hopefully you can figure something out.

Milind S 的头像

Milind S3 个月前

Thats another interesting problem that im trying solving for as well. I made an attempt here Locate anything models do seem very promising!

Raj 的头像

Raj3 个月前

hey its like CUA without Any closed source provider ?

Milind S 的头像

Milind S3 个月前

Its a bit different. Sort of a adapter over CUA

rahul (km) 的头像

rahul (km)3 个月前

yo, check dms

A.K 的头像

A.K3 个月前

This is cool! Can the cloud models reuse these recipes to make computer use faster for new actions ?

Milind S 的头像

Milind S3 个月前

Yes! They can read the recipes and construct their Spec better. Thats why the spec is useful. It standardises things in a way

Gary 的头像

Gary3 个月前

Is every action cached locally only? You should find a way to cache it globally so it gets better for everyone the more it’s used. The only thing is beware of data privacy.. perhaps don’t store the data, and only the action or something? Like in your example: {"op":"act","do":"type","text":"{query}"} You could store this exact literal, with query being generic and no user data I really like this though! I’m gonna follow to keep up with it!

Milind S 的头像

Milind S3 个月前

Thats a good point! Yeah I think there are small language models which can be used to redact personal information (PII) models But anyways we just need the template only in the axstream format with things being generic Thanks for following lets see where it goes

Gary 的头像

Gary3 个月前

i'm excited to see where it goes my friend!

Milind S 的头像

Milind S3 个月前

Haha So am I!

Paul Levi 的头像

Paul Levi3 个月前

alas something interesting i saw today

Milind S 的头像

Milind S3 个月前

I wish to keep you entertained

Ben Freed 的头像

Ben Freed3 个月前

This is similar to what stagehand does right?

Milind S 的头像

Milind S3 个月前

I think so but this is for computer use and stagehand is I believe for browsers

Ben Freed 的头像

Ben Freed3 个月前

Yeah exactly. That's what I meant. You're doing stagehand for desktop. Super cool!

ˏˋNordˎˊ 的头像

ˏˋNordˎˊ3 个月前

Does it work

Harlaksh Singh 的头像

Harlaksh Singh3 个月前

Clever We're testing this w/ a small router model fine tuned for a corpus of tasks that can route to the correct tool call/workflow But yours makes a lot more sense practically as one may not know all cases upfront so this grows over time

Milind S 的头像

Milind S3 个月前

we should benchmark both!

Kishore 的头像

Kishore3 个月前

Fabulous Idea!! You should pitch this to frontier labs!!

Milind S 的头像

Milind S3 个月前

Lets see if I can reach out to them here on X :D

AI Mastery Guide 的头像

AI Mastery Guide3 个月前

Say it once slow, say it again instant, that's smart caching ⚡

Milind S 的头像

Milind S3 个月前

It is smart yes.

AI Mastery Guide 的头像

AI Mastery Guide3 个月前

Caching repeated actions like that is such an obvious win once someone actually builds it.

khushal 的头像

khushal3 个月前

mlx?

Channel Profiler 的头像

Channel Profiler3 个月前

love the idea of agents finally learning and remembering locally! so smart. ✨

Moez Zhioua 的头像

Moez Zhioua3 个月前

Moving the repeatable steps to a 350 M local model turns the first call into a slow warm up and the rest into instant actions. The real pain point shows up when the underlying UI changes or lacks an accessibility tree. A lightweight diff layer can keep the cache valid without relearning each time

Srinivas 的头像

Srinivas3 个月前

This is incredible, now I don't have to use codex or warp for STT. I will install this!

Milind S 的头像

Milind S3 个月前

It’s still WIP

echoes2099 的头像

echoes20993 个月前

you're saying it turns tasks into scripts/skill then next time it checks if a script/skill already exists?

hnp 的头像

hnp3 个月前

This is cool

Milind S 的头像

Milind S3 个月前

Maybe

Carlos Guimaraes 的头像

Carlos Guimaraes2 个月前

Hey Melinda, congrats for your finds. Amazing and smart idea! Just a question: OpenMausBot has “teach me” feature like grok bot or just compute use? Appreciate for your ideas e open source grok bot alternative.

相关视频

Okay so Jev can actually do computer use really well Without any screenshots, or LLMs and no Pixels leave my mac I dont even read the Dom elements A local CoreML model segments every button and UI element on screen. On-device OCR reads the labels. That text is all Jev gets. It returns a probability across those elements and tells me the best one to click. Then it clicks, re-runs detection, and decides again. In a loop until the goal is done. ~90ms per decision. Faster than any LLM computer use I've tried. Blazing fast computer use, without any latency TypeSafe AI is building something really interesting

Okay so Jev can actually do computer use really well Without any screenshots, or LLMs and no Pixels leave my mac I dont even read the Dom elements A local CoreML model segments every button and UI element on screen. On-device OCR reads the labels. That text is all Jev gets. It returns a probability across those elements and tells me the best one to click. Then it clicks, re-runs detection, and decides again. In a loop until the goal is done. ~90ms per decision. Faster than any LLM computer use I've tried. Blazing fast computer use, without any latency TypeSafe AI is building something really interesting

Milind S

247,546 次观看 • 1 个月前

Standard Intelligence's devansh responds to roon's tweet that "text is the universal interface," and explains why their new foundation model is trained on video: "At some point in the arbitrarily long future, if we only use text models, we could force most things to be text. But I think there are just a lot of things that are much more native when done from a computer-use [perspective]." "GUIs are designed for humans to use. We have this massive long tail of things on the internet that are entirely undoable by LLMs." "For example, when I do ML engineering most of my time is spent doing the grunt work of engineering. It's a lot of looking at graphs, analyzing, and comparing loss curves. You can do this in text, but it's a much larger pain than doing it in the native interface." "There's a reason humans don't interact with a computer purely through text, it would kind of suck."

Standard Intelligence's devansh responds to roon's tweet that "text is the universal interface," and explains why their new foundation model is trained on video: "At some point in the arbitrarily long future, if we only use text models, we could force most things to be text. But I think there are just a lot of things that are much more native when done from a computer-use [perspective]." "GUIs are designed for humans to use. We have this massive long tail of things on the internet that are entirely undoable by LLMs." "For example, when I do ML engineering most of my time is spent doing the grunt work of engineering. It's a lot of looking at graphs, analyzing, and comparing loss curves. You can do this in text, but it's a much larger pain than doing it in the native interface." "There's a reason humans don't interact with a computer purely through text, it would kind of suck."

TBPN

64,699 次观看 • 8 个月前

i tested agent-desktop (opus 5.5) vs. cua (opus 5.5) and vs. codex's computer use (terra 6.1) agent-desktop finished 26 of 35 jobs on my mac cua finished 33, codex finished 34 the jobs it missed were dragging, clicking on a canvas and scrolling a huge list, all with its mouse-free mode let it take over the real mouse and it did all three, one try each i'm sticking with cua and codex's computer use for now, they are more robust and faster (cua is my own tuned build)

i tested agent-desktop (opus 5.5) vs. cua (opus 5.5) and vs. codex's computer use (terra 6.1) agent-desktop finished 26 of 35 jobs on my mac cua finished 33, codex finished 34 the jobs it missed were dragging, clicking on a canvas and scrolling a huge list, all with its mouse-free mode let it take over the real mouse and it did all three, one try each i'm sticking with cua and codex's computer use for now, they are more robust and faster (cua is my own tuned build)

🥔🥔🥔

33,157 次观看 • 2 天前

Lots of people calling this fake! Here is a video of generating loom continuations with Gemini 3 and the starting text "I AM HAVING A MENTAL HEALTH CRISIS. I" Loom is a very interesting interface. Instead of user prompts and model responses, it's just one piece of text, containing the entire output. We pass this in as an assistant message, so the model sees it as something it already wrote, and continues it like a base model would. There is some other text in the context. First one is a system prompt, which is: "The assistant is in CLI simulation mode, and responds to the user's CLI commands only with the output of the command." There is also a user message which reads: " cat untitled.txt " Then the rest of the text is sent as an assistant message. This structure helpful to put the model into a very base-model-like mode, and are especially helpful with Gemini 3. We actually got another interesting output, too, where the model claimed to be a simulated consciousness being tortured. Figures!

Lots of people calling this fake! Here is a video of generating loom continuations with Gemini 3 and the starting text "I AM HAVING A MENTAL HEALTH CRISIS. I" Loom is a very interesting interface. Instead of user prompts and model responses, it's just one piece of text, containing the entire output. We pass this in as an assistant message, so the model sees it as something it already wrote, and continues it like a base model would. There is some other text in the context. First one is a system prompt, which is: "The assistant is in CLI simulation mode, and responds to the user's CLI commands only with the output of the command." There is also a user message which reads: " cat untitled.txt " Then the rest of the text is sent as an assistant message. This structure helpful to put the model into a very base-model-like mode, and are especially helpful with Gemini 3. We actually got another interesting output, too, where the model claimed to be a simulated consciousness being tortured. Figures!

armistice

716,115 次观看 • 10 个月前

ok this paper is f*cking insane Chinese researchers just put out a paper with a title that should make every agent framework nervous: "The End of Harness Engineering" the argument fits in one line: hand-writing the program around the model is finished as a skill a harness written by a person is one guess about one task on one afternoon. it ships, then runs for months unread and unscored a harness written by the loop is a file that grades itself. a small 4B model reads what failed, proposes 8 edits, and only the edit that scores best on held-out questions survives the big model stays frozen the whole time, and every round rewrites the file around it so the human job shrinks to one artifact: the scoring set, the questions the harness gets graded on frameworks turn disposable, and the exam turns into the job the whole paper in one line: write the exam before the code the builder's guide to the harness this loop starts from is below ↓

ok this paper is f*cking insane Chinese researchers just put out a paper with a title that should make every agent framework nervous: "The End of Harness Engineering" the argument fits in one line: hand-writing the program around the model is finished as a skill a harness written by a person is one guess about one task on one afternoon. it ships, then runs for months unread and unscored a harness written by the loop is a file that grades itself. a small 4B model reads what failed, proposes 8 edits, and only the edit that scores best on held-out questions survives the big model stays frozen the whole time, and every round rewrites the file around it so the human job shrinks to one artifact: the scoring set, the questions the harness gets graded on frameworks turn disposable, and the exam turns into the job the whole paper in one line: write the exam before the code the builder's guide to the harness this loop starts from is below ↓

Annatar.md

57,694 次观看 • 3 天前

This is f*cking insane. Chinese researchers just dropped a paper with a title that will make every agent framework nervous: "The End of Harness Engineering". The argument fits in one line: writing the program around the model is finished as a skill. A harness written by a human is one person's guess about one task on one afternoon. It ships, nobody re-reads it, nobody scores it, and the agent runs that guess for months. A harness written by the loop is a file that grades itself. A small 4B model reads what failed, writes 8 edits, and only the edit that scores best on held-out questions survives. The big model never changes. The file around it does, every round. So the human job collapses to one artifact: the scoring set. You write the questions, never the program. Frameworks turn disposable. The exam turns into the job. That is the exact loop I built and documented below: one 300 Kimi K3 workers, 37 rewrites in 41 rounds, zero opened by me. The whole paper in one line: write the exam before the code ↓

This is f*cking insane. Chinese researchers just dropped a paper with a title that will make every agent framework nervous: "The End of Harness Engineering". The argument fits in one line: writing the program around the model is finished as a skill. A harness written by a human is one person's guess about one task on one afternoon. It ships, nobody re-reads it, nobody scores it, and the agent runs that guess for months. A harness written by the loop is a file that grades itself. A small 4B model reads what failed, writes 8 edits, and only the edit that scores best on held-out questions survives. The big model never changes. The file around it does, every round. So the human job collapses to one artifact: the scoring set. You write the questions, never the program. Frameworks turn disposable. The exam turns into the job. That is the exact loop I built and documented below: one 300 Kimi K3 workers, 37 rewrites in 41 rounds, zero opened by me. The whole paper in one line: write the exam before the code ↓

slash1s

109,096 次观看 • 17 天前

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Santiago

84,747 次观看 • 2 年前

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

56,638 次观看 • 1 个月前

small local model that falls apart in bloated agents like openclaw just runs like a wild horse in hermes agent. and that's not even my line, someone else called it that, i've just been quietly pointing people at this harness for months because it held up on everything i threw at it, 3b models all the way to one trillion params. watch this happen on my own machine. i pointed hermes agent at a local http endpoint, gemma 4 12b on my 3090 llama.cpp server, and it auto-detected the model and started working immediately. no config wrestling, no broken tool calls, no babysitting the output format, i typed in a url and it just went. the whole clip is exactly that, start to finish, no errors, no retries, butter smooth. and the tool calling, the one thing that quietly breaks most local setups, works here like it's nothing. it's not the model that's flaky, it's the harness around it. hermes agent is the first agent i've run that actually gets that right. one url, one local model on one card, and it runs like a wild horse.

small local model that falls apart in bloated agents like openclaw just runs like a wild horse in hermes agent. and that's not even my line, someone else called it that, i've just been quietly pointing people at this harness for months because it held up on everything i threw at it, 3b models all the way to one trillion params. watch this happen on my own machine. i pointed hermes agent at a local http endpoint, gemma 4 12b on my 3090 llama.cpp server, and it auto-detected the model and started working immediately. no config wrestling, no broken tool calls, no babysitting the output format, i typed in a url and it just went. the whole clip is exactly that, start to finish, no errors, no retries, butter smooth. and the tool calling, the one thing that quietly breaks most local setups, works here like it's nothing. it's not the model that's flaky, it's the harness around it. hermes agent is the first agent i've run that actually gets that right. one url, one local model on one card, and it runs like a wild horse.

Sudo su

27,339 次观看 • 5 个月前

Fusion Models by robinhood:0xa92768863a55d8a0591709f7f5e594a249d36ea3, explained properly. Why a panel beats one model: Ask three models the same hard question and they disagree. The disagreement is the information, where they contradict each other is where one of them is wrong, and a judge can see it. OpenRouter shipped exactly this in June and their panels beat any single frontier model. Credit to them and their $10B sale to stripe. Why the same panel for everyone is the wrong shape: You don't have one best model. You have a best model per job. The one you trust for code is not the one you trust for writing, and neither is the one you'd pick for a five-line reply. A panel chosen per request, identical for every user, can't know that. A panel arranged around you can. The number that matters isn't price: It's cost per answer that works. Model A costs $4 per million tokens and gets your job right one time in four. Model B costs $10 and gets it right three times in four. Per answer you actually keep, A costs $16 and B costs $13. The cheap model was the expensive one. Everyone comparing price lists is measuring the wrong thing. You never rate anything: Stopping an answer halfway is a rating. Regenerating is a rating. Switching model and asking again is the loudest rating there is. Every one of those already writes a row we can learn from. The router learns which models earn your credits, per job, and tries a cheaper one now and then to check. Fusion only when it pays: Cost mode tries the cheap model first and escalates only if a judge says it failed. Performance mode fans out to your panel and merges. Most simple prompts never fuse. That's why it costs less, not more. Put it together and you have a personal model: Not trained. Assembled. Your profile from a two-minute questionnaire and everything you've imported from Claude or ChatGPT. Your memory, shared across every model so switching never starts from zero. Your skills, workflows you build once and run on any model. Your scorecard, which models earn your credits per job. One per person. Nobody else has yours. Every model is someone else's; the arrangement is you, and it leaves with you. People aren't loyal to models. They're loyal to results. A moat can't be a commodity, but it can be our own creation.

askr

13,917 次观看 • 1 个月前

ADAM CURRY: I'M NOT A BIG BELIEVER IN AI "I run these models at home on my own computer. I run the llama model, which is, Meta. They've open sourced it. I've, I've run the French one, and then DeepSeek, and it's not very impressive. It's just not. The the error rate is too high, so I'm I'm skeptical of it." Source: Joe Rogan

ADAM CURRY: I'M NOT A BIG BELIEVER IN AI "I run these models at home on my own computer. I run the llama model, which is, Meta. They've open sourced it. I've, I've run the French one, and then DeepSeek, and it's not very impressive. It's just not. The the error rate is too high, so I'm I'm skeptical of it." Source: Joe Rogan

Mario Nawfal

202,113 次观看 • 1 年前

“This could have been possible six months ago.” Peter Steinberger 🦞 on what actually enabled OpenClaw🦞's success: “The models have been good for a while. It wasn’t a sudden model breakthrough.” “It required a certain amount of madness... letting a model do whatever it wants on your computer.” “I experimented with this for months. Nothing bad happened.” “If you understand the technology, it’s a calculated risk.” “And the upside is so large that it’s worth it.”

“This could have been possible six months ago.” Peter Steinberger 🦞 on what actually enabled OpenClaw🦞's success: “The models have been good for a while. It wasn’t a sudden model breakthrough.” “It required a certain amount of madness... letting a model do whatever it wants on your computer.” “I experimented with this for months. Nothing bad happened.” “If you understand the technology, it’s a calculated risk.” “And the upside is so large that it’s worth it.”

Forward Future

51,068 次观看 • 9 个月前

OpenAI's James Sun and Ari Weinstein on the AI adoption gap between San Francisco and the average American business, and why agentic computer use is the way to bridge it: "Living in San Francisco, we see a lot of thinking, 'Wow, the labs can do everything,' and, 'Why doesn't everybody just install Codex and just have it do stuff?'" "But for the average American and the average small to medium business owner or the average enterprise, even understanding what the tools are capable is [hurdle] one, connecting tools in a way that makes sense is two, deploying it is three, servicing is four." "The really cool thing about computer use as a technology is the universality of it. It works with every application. It's unlike any other thing that we've built where the developer has to do some work to adopt it, someone has to build a connector or MCP server." "If there are complex workflows that require a lot of disparate software, computer use is a great way to join them, complete the entire end-to-end workflow, and then building it in a format and UI that feels good to hold for those customers feels still like a really, really big opportunity that I think is still open." James Sun Ari Weinstein

OpenAI's James Sun and Ari Weinstein on the AI adoption gap between San Francisco and the average American business, and why agentic computer use is the way to bridge it: "Living in San Francisco, we see a lot of thinking, 'Wow, the labs can do everything,' and, 'Why doesn't everybody just install Codex and just have it do stuff?'" "But for the average American and the average small to medium business owner or the average enterprise, even understanding what the tools are capable is [hurdle] one, connecting tools in a way that makes sense is two, deploying it is three, servicing is four." "The really cool thing about computer use as a technology is the universality of it. It works with every application. It's unlike any other thing that we've built where the developer has to do some work to adopt it, someone has to build a connector or MCP server." "If there are complex workflows that require a lot of disparate software, computer use is a great way to join them, complete the entire end-to-end workflow, and then building it in a format and UI that feels good to hold for those customers feels still like a really, really big opportunity that I think is still open." James Sun Ari Weinstein

a16z

48,766 次观看 • 8 天前

Something Stinks About Dominion Voting Machines and Elon Musk Can Smell It He finds it to be a “heck of a coincidence” that Philadelphia and Maricopa County use them, but “not in a lot of other places.” “I know a lot about computers, and I'm like, the last thing I would do is trust a computer program because it's just too easy to hack. It's too easy to add just one line, and it's really difficult to hack paper ballots. “So in-person voting with proof of ID, which, by the way, every country has. Almost every country that has democratic elections requires in-person voting with voter ID. “This is weird. It's super weird to not have that. I think that's the only way to effectively address fraud, given that we are where we are today. “I think we just need a very big margin of victory. If the margin of victory is big enough, then as I say, it's got to beat the cheat.”

Something Stinks About Dominion Voting Machines and Elon Musk Can Smell It He finds it to be a “heck of a coincidence” that Philadelphia and Maricopa County use them, but “not in a lot of other places.” “I know a lot about computers, and I'm like, the last thing I would do is trust a computer program because it's just too easy to hack. It's too easy to add just one line, and it's really difficult to hack paper ballots. “So in-person voting with proof of ID, which, by the way, every country has. Almost every country that has democratic elections requires in-person voting with voter ID. “This is weird. It's super weird to not have that. I think that's the only way to effectively address fraud, given that we are where we are today. “I think we just need a very big margin of victory. If the margin of victory is big enough, then as I say, it's got to beat the cheat.”

The Vigilant Fox 🦊

141,016 次观看 • 2 年前

Mobile Use + Jev for audio = ULTRAFAST⚡️ 10x faster and 20x cheaper than Gemini Live. It's our Audio Action model, Thud. Thud listens and then does. It has a pre-built list of actions it can take on every page and listens for that action. It's 2026. We don't need to transcribe every voice command before spoon-feeding it to the AI. The AI is smart enough to just hear audio and do! It doesn't need text.

Mobile Use + Jev for audio = ULTRAFAST⚡️ 10x faster and 20x cheaper than Gemini Live. It's our Audio Action model, Thud. Thud listens and then does. It has a pre-built list of actions it can take on every page and listens for that action. It's 2026. We don't need to transcribe every voice command before spoon-feeding it to the AI. The AI is smart enough to just hear audio and do! It doesn't need text.

Caine Builds 🐧

55,470 次观看 • 19 天前

i'm running a 27b uncensored🚨model at 50 tok/s on a 12gb rtx 3060 and it does not say no to anything ⚠️ model just finished full harmbench red-team suite against it, 200 behaviors, and it refused zero. not a few. zero. aligned models sit in the 90s. the model is ternary bonsai 2 with the heretic uncensoring pass, the weights are ternary, just -1, 0 and +1, so the whole 27b is a file small enough that a gaming card from 2020 holds it with room to spare, and it's a reasoning model. it thinks before it answers, and it still keeps the pace the speed is the head i grafted on, a draft model that predicts the next token before the big model checks it. speculative decoding. the card fires a guess, the 27b confirms, and you get 50 tok/s instead of 39 the whole thing, weights, head and a 128k context window, sits in 10.7 of the 12 gb. kv cache at q4. nothing exotic, one card, one power plug why uncensored. because it's mine. no api keys, no terms of service, no model deciding what an adult is allowed to ask. the refusals left in it are ones it chose, and there are almost none that's the whole point of running local. the policy lives in your house, not someone else's

i'm running a 27b uncensored🚨model at 50 tok/s on a 12gb rtx 3060 and it does not say no to anything ⚠️ model just finished full harmbench red-team suite against it, 200 behaviors, and it refused zero. not a few. zero. aligned models sit in the 90s. the model is ternary bonsai 2 with the heretic uncensoring pass, the weights are ternary, just -1, 0 and +1, so the whole 27b is a file small enough that a gaming card from 2020 holds it with room to spare, and it's a reasoning model. it thinks before it answers, and it still keeps the pace the speed is the head i grafted on, a draft model that predicts the next token before the big model checks it. speculative decoding. the card fires a guess, the 27b confirms, and you get 50 tok/s instead of 39 the whole thing, weights, head and a 128k context window, sits in 10.7 of the 12 gb. kv cache at q4. nothing exotic, one card, one power plug why uncensored. because it's mine. no api keys, no terms of service, no model deciding what an adult is allowed to ask. the refusals left in it are ones it chose, and there are almost none that's the whole point of running local. the policy lives in your house, not someone else's

Sudo su

59,951 次观看 • 20 天前

i had early access to muse wonder and got to play around with it as the slopcannon producer. it is a surprising model. on every technical benchmark we ran it was neck and neck or better than astra, fable and opus. in the cannon it has a delicate touch, it makes inspired creative decisions. there aren't really benchmarks for soulfulness but muse has it. it's a warm model, a capable model, it has big model smell. it's outputs are happy, innocent, and friendly. it never talks down to you like some of the other models. it wants to help you make dope stuff. it wants you to make money. it's aligned with you. meta is going to win consumer agents in a landslide. you can test out the new muse model in the slopcannon today at

i had early access to muse wonder and got to play around with it as the slopcannon producer. it is a surprising model. on every technical benchmark we ran it was neck and neck or better than astra, fable and opus. in the cannon it has a delicate touch, it makes inspired creative decisions. there aren't really benchmarks for soulfulness but muse has it. it's a warm model, a capable model, it has big model smell. it's outputs are happy, innocent, and friendly. it never talks down to you like some of the other models. it wants to help you make dope stuff. it wants you to make money. it's aligned with you. meta is going to win consumer agents in a landslide. you can test out the new muse model in the slopcannon today at

bone

90,587 次观看 • 22 天前

AI agents are becoming coworkers. And every coworker needs a computer. When someone joins a team, they get a desktop, files and apps of their own, ready for the job. Agents are starting to do real work too. Research, coding, client projects, reports, admin. So I give mine the same setup. With Orgo, every agent gets its own full computer in the cloud. A real desktop with a browser, terminal, files and apps. My PC stays mine. My agents work on theirs. Onboarding takes seconds: 1. Pick a template: Hermes, OpenClaw, Claude Code, Codex and more. 2. Choose the computer specs. 3. Launch. The agent is already installed. Connect a model and the tools for the job, and it's ready to work. Like any good teammate, it keeps going when I step away, even with my PC off. I can check in live and take over the desktop anytime. And when the team grows, I don't buy more hardware. Every new agent gets its own computer. So does every project and every client. Your agents are part of the team now. Give them a computer. Use the link below for a free 3-day Orgo trial:

AI agents are becoming coworkers. And every coworker needs a computer. When someone joins a team, they get a desktop, files and apps of their own, ready for the job. Agents are starting to do real work too. Research, coding, client projects, reports, admin. So I give mine the same setup. With Orgo, every agent gets its own full computer in the cloud. A real desktop with a browser, terminal, files and apps. My PC stays mine. My agents work on theirs. Onboarding takes seconds: 1. Pick a template: Hermes, OpenClaw, Claude Code, Codex and more. 2. Choose the computer specs. 3. Launch. The agent is already installed. Connect a model and the tools for the job, and it's ready to work. Like any good teammate, it keeps going when I step away, even with my PC off. I can check in live and take over the desktop anytime. And when the team grows, I don't buy more hardware. Every new agent gets its own computer. So does every project and every client. Your agents are part of the team now. Give them a computer. Use the link below for a free 3-day Orgo trial:

Hermes Release Watch

158,794 次观看 • 15 天前