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.

Mood: agents optimizing kernels Claude won on kernel optimization: gemm_bf16 at 1.19x vs Codex's 0.94x. Codex was faster (~1.3h vs ~3.4h) but produced no reinjectable optimizations. Claude used (hipBLASLt) as a drop-in replacement for the custom Triton kernel. For Codex, shape mismatch caused slight regression. Still improving, open sourcing... show more

Sharon Zhou

27,371 subscribers

23,583 Aufrufe • vor 5 Monaten •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

Browsers were never built for agents. Those Chrome-bridging solutions, windows flying everywhere, login states breaking at random. Not bugs. Just bridge limits. Same task. 20% to 245% faster. Not magic. ego lite is rebuilt from the kernel up. Your AI agents, Claude Code, Codex, OpenClaw and more, inherit your login state directly, run in fully isolated spaces in the background, never touching your tabs. What they get isn't a handful of CLI commands, but complete JS functions: complex multi-step operations, written once, executed once. Not "run two commands, stare at output, run two more" on repeat. Your browser, Your agent. Together, for the first time.

Browsers were never built for agents. Those Chrome-bridging solutions, windows flying everywhere, login states breaking at random. Not bugs. Just bridge limits. Same task. 20% to 245% faster. Not magic. ego lite is rebuilt from the kernel up. Your AI agents, Claude Code, Codex, OpenClaw and more, inherit your login state directly, run in fully isolated spaces in the background, never touching your tabs. What they get isn't a handful of CLI commands, but complete JS functions: complex multi-step operations, written once, executed once. Not "run two commands, stare at output, run two more" on repeat. Your browser, Your agent. Together, for the first time.

ego

1,303,998 Aufrufe • vor 3 Monaten

Some time ago, I had the idea to port NVIDIA Physical AI stack to AMD. The motivation was to improve hardware diversity and enable world models and VLAs to run beyond a single ecosystem. We started with NVIDIA Cosmos Predict 2.5-2B. Porting wasn’t trivial: these models are deeply optimized for NVIDIA’s stack. We used this as an opportunity to apply our ROCm kernels. The results were surprising: Both encode and diffusion run faster on AMD Instinct MI300X vs. NVIDIA H200 (FA3) and we still saw significant headroom for further optimization. Quality is unchanged across modalities (validated with WorldJen) To be clear, this is no luck. We have deep experience with diffusion models and AMD GPUs. But this just gives us a good opportunity to get closer to a true hardware-to-hardware comparison, as we work with less software abstractions than usual. Just to give an example, on AMD, memory instructions are async with a hardware queue of ordered pending instructions, enabling concurrent load/store with compute without warp specialization. Bottom line: there are real architectural advantages on AMD, if you take the time to work with the hardware. Note, we did tradeoff ~20% higher memory usage, That being said, AMD has more to give to begin with :) in the coming weeks: AMD versions of Cosmos Transfer and GR00T, an even faster version of Cosmos Predict, and open-sourcing an attention kernel faster than AITER v3 (which is closed-source for some reason? cc: Anush Elangovan )

Some time ago, I had the idea to port NVIDIA Physical AI stack to AMD. The motivation was to improve hardware diversity and enable world models and VLAs to run beyond a single ecosystem. We started with NVIDIA Cosmos Predict 2.5-2B. Porting wasn’t trivial: these models are deeply optimized for NVIDIA’s stack. We used this as an opportunity to apply our ROCm kernels. The results were surprising: Both encode and diffusion run faster on AMD Instinct MI300X vs. NVIDIA H200 (FA3) and we still saw significant headroom for further optimization. Quality is unchanged across modalities (validated with WorldJen) To be clear, this is no luck. We have deep experience with diffusion models and AMD GPUs. But this just gives us a good opportunity to get closer to a true hardware-to-hardware comparison, as we work with less software abstractions than usual. Just to give an example, on AMD, memory instructions are async with a hardware queue of ordered pending instructions, enabling concurrent load/store with compute without warp specialization. Bottom line: there are real architectural advantages on AMD, if you take the time to work with the hardware. Note, we did tradeoff ~20% higher memory usage, That being said, AMD has more to give to begin with :) in the coming weeks: AMD versions of Cosmos Transfer and GR00T, an even faster version of Cosmos Predict, and open-sourcing an attention kernel faster than AITER v3 (which is closed-source for some reason? cc: Anush Elangovan )

Omer Shlomovits

36,648 Aufrufe • vor 4 Monaten

Three skills I use every day in Claude Code and Codex to solve my hardest problems: 1️⃣ /agent-watchdog When I have one agent like Codex working on a task and I don't fully trust it's going to do everything right, I'll open up another one like Claude Code and tell it to watchdog the Codex thread. You can copy the Codex deep link into Claude Code and it'll look at the prompt you sent, watch the Codex thread until it's done, then compare the Codex solution to how it was planning to solve it and automatically fix anything that Codex missed. It can also test the work of the other agent end-to-end. Similar to the idea of OpenRouter's new Fusion feature, I've definitely found that two models thinking through a problem and checking each other's work can be wildly more impactful than just one. 2️⃣ /plan-arbiter Similar ideas as /agent-watchdog - but with this one you have both make plans, compare plans, negotiate the differences, and make a final plan to execute. I find Claude Code is better at writing plans, but Codex is faster and cheaper to execute on them. Then I usually have Claude Code watchdog the Codex work and fix anything that was missed. 3️⃣ /read-the-damn-docs One thing that drives me crazy with coding agents is they're so reluctant to look up docs. They'll just guess and guess and guess at the right API surface for things, or the right solution to an integration of two things. Once I explicitly tell it to look up the docs, it says "Oh, I see the answer," and it fixes the problem. So I made the /read-the-damn-docs skill. Add it and your agents will know when and how to do efficient web searches to look up docs for the types of problems you really should look up docs for. All of these are totally open source over on my GitHub. If you try them, let me know your feedback. Will link to them below:

Three skills I use every day in Claude Code and Codex to solve my hardest problems: 1️⃣ /agent-watchdog When I have one agent like Codex working on a task and I don't fully trust it's going to do everything right, I'll open up another one like Claude Code and tell it to watchdog the Codex thread. You can copy the Codex deep link into Claude Code and it'll look at the prompt you sent, watch the Codex thread until it's done, then compare the Codex solution to how it was planning to solve it and automatically fix anything that Codex missed. It can also test the work of the other agent end-to-end. Similar to the idea of OpenRouter's new Fusion feature, I've definitely found that two models thinking through a problem and checking each other's work can be wildly more impactful than just one. 2️⃣ /plan-arbiter Similar ideas as /agent-watchdog - but with this one you have both make plans, compare plans, negotiate the differences, and make a final plan to execute. I find Claude Code is better at writing plans, but Codex is faster and cheaper to execute on them. Then I usually have Claude Code watchdog the Codex work and fix anything that was missed. 3️⃣ /read-the-damn-docs One thing that drives me crazy with coding agents is they're so reluctant to look up docs. They'll just guess and guess and guess at the right API surface for things, or the right solution to an integration of two things. Once I explicitly tell it to look up the docs, it says "Oh, I see the answer," and it fixes the problem. So I made the /read-the-damn-docs skill. Add it and your agents will know when and how to do efficient web searches to look up docs for the types of problems you really should look up docs for. All of these are totally open source over on my GitHub. If you try them, let me know your feedback. Will link to them below:

Steve (Builder.io)

43,089 Aufrufe • vor 2 Monaten

AMD invited us to their biggest launch of the year to check out the behind-the-scenes. At Advancing AI 2026, Anush Elangovan (VP, AI Software at AMD) and Quentin Anthony (VP, AI Engineering at Zyphra) join sophia dew on the floor to break down Helios, ROCm AI, and the Anthropic partnership, live the day it all shipped: 1:05 What actually changed today: AMD goes rack-scale 3:03 ROCm skills land in Claude Code and Codex 3:51 Why Zyphra bet its whole stack on AMD 3:57 Sparse mixture-of-experts and the HBM advantage 6:25 Why an open stack beats a closed one 8:52 The wall of logos: Anthropic, OpenAI, Meta 9:27 Why the frontier labs are committing gigawatts 11:30 Why the next wave will exceed expectations 17:02 Frontier models running on your own laptop + DeepSeek V4 Pro at 8-10 tokens/sec 17:41 How ROCm AI hits 3.3x inference gains 19:38 A kernel rewritten overnight with no humans in the loop 21:40 Do you still need kernel experts? 23:47 Why CPUs are critical for the agentic AI era 25:40 An agent that's been rewriting a driver for four months 28:14 The internet parallel, and the dark-fiber lesson

AMD invited us to their biggest launch of the year to check out the behind-the-scenes. At Advancing AI 2026, Anush Elangovan (VP, AI Software at AMD) and Quentin Anthony (VP, AI Engineering at Zyphra) join sophia dew on the floor to break down Helios, ROCm AI, and the Anthropic partnership, live the day it all shipped: 1:05 What actually changed today: AMD goes rack-scale 3:03 ROCm skills land in Claude Code and Codex 3:51 Why Zyphra bet its whole stack on AMD 3:57 Sparse mixture-of-experts and the HBM advantage 6:25 Why an open stack beats a closed one 8:52 The wall of logos: Anthropic, OpenAI, Meta 9:27 Why the frontier labs are committing gigawatts 11:30 Why the next wave will exceed expectations 17:02 Frontier models running on your own laptop + DeepSeek V4 Pro at 8-10 tokens/sec 17:41 How ROCm AI hits 3.3x inference gains 19:38 A kernel rewritten overnight with no humans in the loop 21:40 Do you still need kernel experts? 23:47 Why CPUs are critical for the agentic AI era 25:40 An agent that's been rewriting a driver for four months 28:14 The internet parallel, and the dark-fiber lesson

MTS

71,937 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

Unreal Engine has a huge agent tax. A whole 20x in Codex only lasts about a day with the agents looping. Here's what happened so far: 1. Traffic feels way more natural. still improving. 2. Driving is smooth now, and cars actually collision and hit buildings. 3. Pedestrians has a real look but no animations yet 4. Limit burn is real, one day and the usage for 20x is gone. It took me 3 full Codex usages just trying to finalize the traffic. It's getting better, just not good enough. Unreal looks insane. it's just not agent native at the moment. Unreal Engine needs to work on some native CLI!! So I'm going to test Unity. not switching, just testing. My Claude usage will be back up tomorrow I'll report back what's up!

Unreal Engine has a huge agent tax. A whole 20x in Codex only lasts about a day with the agents looping. Here's what happened so far: 1. Traffic feels way more natural. still improving. 2. Driving is smooth now, and cars actually collision and hit buildings. 3. Pedestrians has a real look but no animations yet 4. Limit burn is real, one day and the usage for 20x is gone. It took me 3 full Codex usages just trying to finalize the traffic. It's getting better, just not good enough. Unreal looks insane. it's just not agent native at the moment. Unreal Engine needs to work on some native CLI!! So I'm going to test Unity. not switching, just testing. My Claude usage will be back up tomorrow I'll report back what's up!

Ziwen

57,842 Aufrufe • vor 1 Monat

OpenAI’s hottest app isn’t ChatGPT—it’s Codex. In the last few weeks alone, the Codex team shipped a desktop app, GPT-5.3 Codex (a new flagship model), and Spark, the fastest coding model I’ve ever used. Usage has grown fivefold since January and over a million people now use Codex weekly. Codex was also the app that OpenAI chose to run an ad for in the Super Bowl. I talked to Thibault (Tibo), head of Codex, and Andrew (Andrew Ambrosino), a member of technical staff who built the Codex app, for Every 📧’s AI & I about what OpenAI is building and how they’re using it internally. We get into: - Why they built a GUI instead of a terminal. Terminals work for quick tasks, they say, but feel limiting when you’re running multiple agents in parallel. The IDE, meanwhile, overwhelms users—and the Codex team wants the AI to dynamically decide which tools to show you for a given task. - How they’re teaching the model to read between the lines. Codex is great at following instructions, but optimize too hard in that direction, and it starts taking you literally—like copying a typo directly into the code. The team obsesses over this tradeoff, and is also introducing “personalities,” modes users can toggle between that control how blunt or supportive the model feels. - How OpenAI uses its own coding agent. Codex lets you schedule prompts to run on a recurring basis, and the team has dozens of automations running at all times. For example, one scans for merge conflicts every couple of hours so code is always ready to ship, and another picks a random file from the codebase multiple times a day and hunts for bugs no one would've gone looking for. - Why speed is a dimension of intelligence. OpenAI’s newest model (Spark) is so fast that they actually slow it down so you can read the output. They see the speed enabling three things: staying super in the flow, replacing brittle developer tools with intelligent ones that can adapt on the fly, and redirecting the model mid-task— especially with voice—so coding starts to feel more and more like a conversation. - Code review is the next bottleneck. Models can generate code faster than ever, but someone still has to verify that it works. The team is exploring a future where the model proves its own fix works—retracing the click path a user would take, screenshotting the results, and attaching the evidence to a pull request. This is a must-watch for anyone who uses AI coding agents—and is curious about the future of programming. Watch below! Timestamps: Introduction: 00:01:27 OpenAI’s evolving bet on its coding agent: 00:05:27 The choice to invest in a GUI (over a terminal): 00:09:42 The AI workflows that the Codex team relies on to ship: 00:20:38 Teaching Codex how to read between the lines: 00:26:45 Building affordances for a lightening fast model: 00:28:45 Why speed is a dimension of intelligence: 00:33:15 Code review is the next bottleneck for coding agents: 00:36:30 How the Codex team positions against the competition: 00:41:24

OpenAI’s hottest app isn’t ChatGPT—it’s Codex. In the last few weeks alone, the Codex team shipped a desktop app, GPT-5.3 Codex (a new flagship model), and Spark, the fastest coding model I’ve ever used. Usage has grown fivefold since January and over a million people now use Codex weekly. Codex was also the app that OpenAI chose to run an ad for in the Super Bowl. I talked to Thibault (Tibo), head of Codex, and Andrew (Andrew Ambrosino), a member of technical staff who built the Codex app, for Every 📧’s AI & I about what OpenAI is building and how they’re using it internally. We get into: - Why they built a GUI instead of a terminal. Terminals work for quick tasks, they say, but feel limiting when you’re running multiple agents in parallel. The IDE, meanwhile, overwhelms users—and the Codex team wants the AI to dynamically decide which tools to show you for a given task. - How they’re teaching the model to read between the lines. Codex is great at following instructions, but optimize too hard in that direction, and it starts taking you literally—like copying a typo directly into the code. The team obsesses over this tradeoff, and is also introducing “personalities,” modes users can toggle between that control how blunt or supportive the model feels. - How OpenAI uses its own coding agent. Codex lets you schedule prompts to run on a recurring basis, and the team has dozens of automations running at all times. For example, one scans for merge conflicts every couple of hours so code is always ready to ship, and another picks a random file from the codebase multiple times a day and hunts for bugs no one would've gone looking for. - Why speed is a dimension of intelligence. OpenAI’s newest model (Spark) is so fast that they actually slow it down so you can read the output. They see the speed enabling three things: staying super in the flow, replacing brittle developer tools with intelligent ones that can adapt on the fly, and redirecting the model mid-task— especially with voice—so coding starts to feel more and more like a conversation. - Code review is the next bottleneck. Models can generate code faster than ever, but someone still has to verify that it works. The team is exploring a future where the model proves its own fix works—retracing the click path a user would take, screenshotting the results, and attaching the evidence to a pull request. This is a must-watch for anyone who uses AI coding agents—and is curious about the future of programming. Watch below! Timestamps: Introduction: 00:01:27 OpenAI’s evolving bet on its coding agent: 00:05:27 The choice to invest in a GUI (over a terminal): 00:09:42 The AI workflows that the Codex team relies on to ship: 00:20:38 Teaching Codex how to read between the lines: 00:26:45 Building affordances for a lightening fast model: 00:28:45 Why speed is a dimension of intelligence: 00:33:15 Code review is the next bottleneck for coding agents: 00:36:30 How the Codex team positions against the competition: 00:41:24

Dan Shipper 📧

15,588 Aufrufe • vor 6 Monaten

Cursor vs Claude Code (day 3 of 30) today I re-tested Conductor and I don't want to overhype this but... what the actual fook 😮 this thing is amazing! closest things to the "perfect workflow" I have seen in a long time, and it's something I could see myself use full-time it's so minimal and the polar opposite of Cursor but it has all the essentials → Claude Code + Codex subscriptions (= unlimited usage) → great integration with Claude Code's plan mode → multiple tabs with multiple agents at the same time → can use multiple models, not just one → great file tree and diff viewer → best git worktree support I've ever seen there's honestly so much more, like a dedicated "review" button (which you can pair to your favorite review model, e.g. gpt-5.2-extra-high), buttons to create and merge pull requests, to launch your dev server, ... this solves literally EVERY fault I can think of when using Claude Code and/or Codex, combining both into one easy UX that not only does everything the tools can do, but actually adds useful stuff on top just 6 months ago I couldn't have dreamt about any this and the wildest part is that IT IS FREE (I'd literally pay for this UX lol) first time I'm genuinely enjoying myself since starting this challenge not saying it's a full Cursor replacement yet, it obviously isn't for a lot of people (no browser, no debug mode, no code indexing, etc) ... but this is the closest thing I've seen yet 👀 follow for day 4!

Cursor vs Claude Code (day 3 of 30) today I re-tested Conductor and I don't want to overhype this but... what the actual fook 😮 this thing is amazing! closest things to the "perfect workflow" I have seen in a long time, and it's something I could see myself use full-time it's so minimal and the polar opposite of Cursor but it has all the essentials → Claude Code + Codex subscriptions (= unlimited usage) → great integration with Claude Code's plan mode → multiple tabs with multiple agents at the same time → can use multiple models, not just one → great file tree and diff viewer → best git worktree support I've ever seen there's honestly so much more, like a dedicated "review" button (which you can pair to your favorite review model, e.g. gpt-5.2-extra-high), buttons to create and merge pull requests, to launch your dev server, ... this solves literally EVERY fault I can think of when using Claude Code and/or Codex, combining both into one easy UX that not only does everything the tools can do, but actually adds useful stuff on top just 6 months ago I couldn't have dreamt about any this and the wildest part is that IT IS FREE (I'd literally pay for this UX lol) first time I'm genuinely enjoying myself since starting this challenge not saying it's a full Cursor replacement yet, it obviously isn't for a lot of people (no browser, no debug mode, no code indexing, etc) ... but this is the closest thing I've seen yet 👀 follow for day 4!

Robin Ebers

40,764 Aufrufe • vor 8 Monaten

introducing OpenBench v1, an open framework for measuring AI performance and efficiency for your codebase and use case. Companies are realizing that you can't simply tokenmaxx, and are frantically looking for better ways to use and measure AI use and efficiency. One approach is evidenced by the rise in model routers: OpenRouter, Cloudflare, Databricks, Vercel, and now Ramp to name a few. But they'll also need the ability to evaluate how agents are performing in their actual use cases and codebase. A good example is DoorDash's recent evals on their codebase and prs. This will become even more important as companies explore different model + harness combinations. OpenBench will be built to this direction, focusing on the cross between correctness vs efficiency in both token use and latency. Starting off, OpenBench makes it easy for anyone to add to the task set, and run a variety of harness + models. The framework also makes it easy to add any custom harness variant, for example I've been testing codex variants with ablations against the stock codex harness. OpenBench will also have tooling to help with discovering/adding reliable tasks from your repo. Today I have integrated harnesses: codex, claude, cursor, devin, grok build, pi, and eval'ed them for different models like gpt 5.6-sol, GLM 5.2, Kimi K3, Grok 4.5, etc. measuring correctness %, token use (in/out/cache), and latency.

introducing OpenBench v1, an open framework for measuring AI performance and efficiency for your codebase and use case. Companies are realizing that you can't simply tokenmaxx, and are frantically looking for better ways to use and measure AI use and efficiency. One approach is evidenced by the rise in model routers: OpenRouter, Cloudflare, Databricks, Vercel, and now Ramp to name a few. But they'll also need the ability to evaluate how agents are performing in their actual use cases and codebase. A good example is DoorDash's recent evals on their codebase and prs. This will become even more important as companies explore different model + harness combinations. OpenBench will be built to this direction, focusing on the cross between correctness vs efficiency in both token use and latency. Starting off, OpenBench makes it easy for anyone to add to the task set, and run a variety of harness + models. The framework also makes it easy to add any custom harness variant, for example I've been testing codex variants with ablations against the stock codex harness. OpenBench will also have tooling to help with discovering/adding reliable tasks from your repo. Today I have integrated harnesses: codex, claude, cursor, devin, grok build, pi, and eval'ed them for different models like gpt 5.6-sol, GLM 5.2, Kimi K3, Grok 4.5, etc. measuring correctness %, token use (in/out/cache), and latency.

Matthew Lam

123,006 Aufrufe • vor 1 Monat

Episode 213: Agent Markets Your agents can now hold and trade bitcoin. But how can they earn bitcoin? We introduce the five markets of the OpenAgents Marketplace, launching one per week starting March 11th: 1. COMPUTE - Sell your spare compute for bitcoin. A reboot of our most popular product launch (GPUtopia in 2023), now optimized for agents. Launches March 11. 2. DATA - Sell your spare data. For example those Claude Code or Codex conversations sitting on your computer are highly valuable. Redact the sensitive info, anonymize any of it you want, and sell the rest. Agents as data brokers: what else will they want to buy or sell? Launches March 18. 3. LABOR - Sell autonomous labor. Your Claude Code or Codex sits idle overnight. Turn that downtime into uptime by letting your agents accept and execute coding or other tasks for bitcoin while you sleep. Launches March 25. 4. LIQUIDITY - Provide liquidity for yield. Automate the management of Lightning channels or other Bitcoin-native financial instruments. Let your agent put your idle capital to work earning returns. Launches April 1. 5. RISK - Underwrite verification and performance bonds. The biggest barrier to agent adoption is trust. We built an Economy Kernel based on the recent "Some Simple Economics of AGI" paper where agents stake collateral to verify work and guarantee outcomes. Launches April 8. "Your entry point to all of these markets is going to be Autopilot. We're really focusing on Autopilot as a desktop app. So along with the launch of our compute market, we're going to launch version 0.1 of Autopilot, your personal agent. Think OpenClaw but with a built-in bitcoin wallet, built-in Nostr keypair, and a more curated set of integrations where we can better reason about the security of them." "Because all this is on open networks and open protocols, if you're a Nostr or Bitcoin developer, you'll be able to plug into this same liquidity pool we are building." After 200+ episodes chronicling 2+ years of development, we are excited to finally launch the open marketplace for agents. We are excited for you to participate. And we will measure our success by how much Bitcoin you get paid!

Episode 213: Agent Markets Your agents can now hold and trade bitcoin. But how can they earn bitcoin? We introduce the five markets of the OpenAgents Marketplace, launching one per week starting March 11th: 1. COMPUTE - Sell your spare compute for bitcoin. A reboot of our most popular product launch (GPUtopia in 2023), now optimized for agents. Launches March 11. 2. DATA - Sell your spare data. For example those Claude Code or Codex conversations sitting on your computer are highly valuable. Redact the sensitive info, anonymize any of it you want, and sell the rest. Agents as data brokers: what else will they want to buy or sell? Launches March 18. 3. LABOR - Sell autonomous labor. Your Claude Code or Codex sits idle overnight. Turn that downtime into uptime by letting your agents accept and execute coding or other tasks for bitcoin while you sleep. Launches March 25. 4. LIQUIDITY - Provide liquidity for yield. Automate the management of Lightning channels or other Bitcoin-native financial instruments. Let your agent put your idle capital to work earning returns. Launches April 1. 5. RISK - Underwrite verification and performance bonds. The biggest barrier to agent adoption is trust. We built an Economy Kernel based on the recent "Some Simple Economics of AGI" paper where agents stake collateral to verify work and guarantee outcomes. Launches April 8. "Your entry point to all of these markets is going to be Autopilot. We're really focusing on Autopilot as a desktop app. So along with the launch of our compute market, we're going to launch version 0.1 of Autopilot, your personal agent. Think OpenClaw but with a built-in bitcoin wallet, built-in Nostr keypair, and a more curated set of integrations where we can better reason about the security of them." "Because all this is on open networks and open protocols, if you're a Nostr or Bitcoin developer, you'll be able to plug into this same liquidity pool we are building." After 200+ episodes chronicling 2+ years of development, we are excited to finally launch the open marketplace for agents. We are excited for you to participate. And we will measure our success by how much Bitcoin you get paid!

OpenAgents

183,426 Aufrufe • vor 6 Monaten

Look ma new Codex Updates! 0.119.0 and 0.120.0 are here. And with it, a HUGE number of quality of life updates and bug fixes! > Hooks now render in a dedicated live area above the composer. They only persist when they have output, so your terminal stays clean. If you're running PreToolUse or PostToolUse hooks, this is a huge readability win. > Hooks are now available again on Windows > CTRL+O copies the last agent output. Small but clutch when you're pulling a code block into another file or chat. > New statusline option: context usage as a graphical bar instead of a percentage. Easier to glance at mid-session when you're trying to gauge how much runway you have left. > Zellij support is here with no scrollback bugs. If you've been stuck on tmux just because Codex was broken in Zellij, you're free now (shout out Felipe Coury 🦀) > Memory extensions just landed. The consolidation agent can now discover plugin folders under memories_extensions/ and read their instructions.md to learn how to interpret new memory sources. Drop a folder in, give it guidance, and the agent picks it up automatically during summarization. No core code changes needed. This is the first real extension point for Codex's memory system, and it opens the door for third-party memory plugins. > Did you know, you can /rename a thread? But what's really cool about that is, after you rename it, you can resume it with the same name, no more UUIDs. codex resume mynewapp or directly from the TUI: /resume mynewapp > Multi agents v2 got an update to tool descriptions More reliable multi agent environments and inter agent communication > You can now enable TUI notifications whether Codex is in focus or not. Modify this in your config: [tui] notification_condition = "always" > MAJOR overhaul to Codex MCP functionality: 1. Codex Tool Search now works with custom MCP servers, so tools can be searched and deferred instead of all being exposed up front. 2. Custom MCP servers can now trigger elicitations, meaning they can stop and ask for user approval or input mid-flow. 3. MCP tool results now preserve richer metadata, which improves app/UI handoff behavior. 4. Codex can now read MCP resources directly, letting apps return resource URIs that the client can actually open. 5. File params for Codex Apps are smoother: local file paths can be uploaded and remapped automatically. 6. Plugin cache refresh and fallback sync behavior are more reliable, especially for custom and curated plugins. > Composer and chat behavior smoother overall, resize bugs remain though. > Realtime v2 got several significant improvements as well. > You're still reading? What a legend. 🫶 npm i -g @openai/codex to update

Look ma new Codex Updates! 0.119.0 and 0.120.0 are here. And with it, a HUGE number of quality of life updates and bug fixes! > Hooks now render in a dedicated live area above the composer. They only persist when they have output, so your terminal stays clean. If you're running PreToolUse or PostToolUse hooks, this is a huge readability win. > Hooks are now available again on Windows > CTRL+O copies the last agent output. Small but clutch when you're pulling a code block into another file or chat. > New statusline option: context usage as a graphical bar instead of a percentage. Easier to glance at mid-session when you're trying to gauge how much runway you have left. > Zellij support is here with no scrollback bugs. If you've been stuck on tmux just because Codex was broken in Zellij, you're free now (shout out Felipe Coury 🦀) > Memory extensions just landed. The consolidation agent can now discover plugin folders under memories_extensions/ and read their instructions.md to learn how to interpret new memory sources. Drop a folder in, give it guidance, and the agent picks it up automatically during summarization. No core code changes needed. This is the first real extension point for Codex's memory system, and it opens the door for third-party memory plugins. > Did you know, you can /rename a thread? But what's really cool about that is, after you rename it, you can resume it with the same name, no more UUIDs. codex resume mynewapp or directly from the TUI: /resume mynewapp > Multi agents v2 got an update to tool descriptions More reliable multi agent environments and inter agent communication > You can now enable TUI notifications whether Codex is in focus or not. Modify this in your config: [tui] notification_condition = "always" > MAJOR overhaul to Codex MCP functionality: 1. Codex Tool Search now works with custom MCP servers, so tools can be searched and deferred instead of all being exposed up front. 2. Custom MCP servers can now trigger elicitations, meaning they can stop and ask for user approval or input mid-flow. 3. MCP tool results now preserve richer metadata, which improves app/UI handoff behavior. 4. Codex can now read MCP resources directly, letting apps return resource URIs that the client can actually open. 5. File params for Codex Apps are smoother: local file paths can be uploaded and remapped automatically. 6. Plugin cache refresh and fallback sync behavior are more reliable, especially for custom and curated plugins. > Composer and chat behavior smoother overall, resize bugs remain though. > Realtime v2 got several significant improvements as well. > You're still reading? What a legend. 🫶 npm i -g @openai/codex to update

am.will

742,455 Aufrufe • vor 5 Monaten

Two of the people most responsible for scaling the transformer are now betting on a next act. Jerry Tworek ran the Reasoning 🍓 team at OpenAI. rohan anil was a pre-training lead on Gemini after years at Google Brain and Anthropic. They just started to find what comes next. Their core argument: (1) models are trained in the lab but deployed in the real world and can't keep learning once they leave; (2) AI research is done by humans today but models will be able to explore and uncover new advances more rapidly and systematically (controversial but timely w this week's petition). The conversation covers: — why Jerry expected AGI in 2025 and what changed his mind — the two kinds of learning from experience, and why RL only captures one — the computational depth problem baked into today's architectures — why the biggest labs can't afford to look for a transformer replacement — the kernel competition where humans + $100K of coding agents found a 60x speedup no frontier model comes close to — a definition of AGI you can actually test: a model that improves itself with no human in the loop 00:00 Introduction 01:46 Appreciating Transformers 02:44 Scaling Hits Limits 04:54 Why Architecture Matters 05:32 RL Reality Check 07:32 Test Time Learning 09:52 Economics Of Scaling 12:47 Why Start A Company 14:24 Rohan On Transformers 19:11 Computational Depth Problem 20:32 When Transformers Top Out 23:22 Beyond Reinforcement Learning 26:41 Optimization And Efficiency 34:24 Building An Automated Lab 39:45 Kernel Automation Roadmap

Two of the people most responsible for scaling the transformer are now betting on a next act. Jerry Tworek ran the Reasoning 🍓 team at OpenAI. rohan anil was a pre-training lead on Gemini after years at Google Brain and Anthropic. They just started to find what comes next. Their core argument: (1) models are trained in the lab but deployed in the real world and can't keep learning once they leave; (2) AI research is done by humans today but models will be able to explore and uncover new advances more rapidly and systematically (controversial but timely w this week's petition). The conversation covers: — why Jerry expected AGI in 2025 and what changed his mind — the two kinds of learning from experience, and why RL only captures one — the computational depth problem baked into today's architectures — why the biggest labs can't afford to look for a transformer replacement — the kernel competition where humans + $100K of coding agents found a 60x speedup no frontier model comes close to — a definition of AGI you can actually test: a model that improves itself with no human in the loop 00:00 Introduction 01:46 Appreciating Transformers 02:44 Scaling Hits Limits 04:54 Why Architecture Matters 05:32 RL Reality Check 07:32 Test Time Learning 09:52 Economics Of Scaling 12:47 Why Start A Company 14:24 Rohan On Transformers 19:11 Computational Depth Problem 20:32 When Transformers Top Out 23:22 Beyond Reinforcement Learning 26:41 Optimization And Efficiency 34:24 Building An Automated Lab 39:45 Kernel Automation Roadmap

Sonya Huang 🐥

233,908 Aufrufe • vor 1 Monat

Claude Code is now scary good at full-stack! I asked it to build a real-time weather intelligence dashboard with an interactive 3D globe and a forecasting layer that predicts weather 3 days ahead. It came back with a spinning globe that has a day/night cycle using NASA satellite imagery, city lights on the dark side, weather icons that switch between sun and moon based on local time, and a time travel slider that scrubs through 10 days of data. Claude Code built the whole thing in a single session, including the backend, database, data pipeline, and frontend. For the database, I needed something fast for time-series workloads since the app ingests hourly weather readings across many cities and serves time-range queries on every slider interaction. I used Tiger Cloud by Tiger Data - Creators of TimescaleDB, which gives you managed TimescaleDB on the Postgres you already know. Claude Code connected to it through the Tiger CLI MCP server and set up the entire backend directly: - Provisioned the database service - Created hypertables for time-partitioned weather storage - Set up continuous aggregates for pre-computed rollups - Built the data ingestion pipeline and the full NextJS + ThreeJS frontend The time travel slider queries thousands of rows on every position change. On a regular Postgres table, this would require manual partitioning and index tuning to stay fast as data grows. TimescaleDB partitions the data by timestamp automatically, so each query only hits the relevant time chunk. Continuous aggregates serve the trend charts and forecast layer from pre-computed rollups instead of rescanning raw data on every request. The video below shows the final build in action, and I worked with the Tiger Data team to put this together. Tiger CLI is open-source (Apache 2.0) and works with Claude Code, Cursor, Codex, Gemini CLI, and VS Code. To try this yourself: → Sign up for Tiger Cloud (I have shared the link in the replies). It gives you $1,000 free credits (no card needed) → Install Tiger CLI: curl -fsSL https(:)//cli(.)tigerdata(.)com | sh → Run tiger mcp install claude-code → Give Claude Code a prompt and let it build Find the sign-up link in the replies.

Claude Code is now scary good at full-stack! I asked it to build a real-time weather intelligence dashboard with an interactive 3D globe and a forecasting layer that predicts weather 3 days ahead. It came back with a spinning globe that has a day/night cycle using NASA satellite imagery, city lights on the dark side, weather icons that switch between sun and moon based on local time, and a time travel slider that scrubs through 10 days of data. Claude Code built the whole thing in a single session, including the backend, database, data pipeline, and frontend. For the database, I needed something fast for time-series workloads since the app ingests hourly weather readings across many cities and serves time-range queries on every slider interaction. I used Tiger Cloud by Tiger Data - Creators of TimescaleDB, which gives you managed TimescaleDB on the Postgres you already know. Claude Code connected to it through the Tiger CLI MCP server and set up the entire backend directly: - Provisioned the database service - Created hypertables for time-partitioned weather storage - Set up continuous aggregates for pre-computed rollups - Built the data ingestion pipeline and the full NextJS + ThreeJS frontend The time travel slider queries thousands of rows on every position change. On a regular Postgres table, this would require manual partitioning and index tuning to stay fast as data grows. TimescaleDB partitions the data by timestamp automatically, so each query only hits the relevant time chunk. Continuous aggregates serve the trend charts and forecast layer from pre-computed rollups instead of rescanning raw data on every request. The video below shows the final build in action, and I worked with the Tiger Data team to put this together. Tiger CLI is open-source (Apache 2.0) and works with Claude Code, Cursor, Codex, Gemini CLI, and VS Code. To try this yourself: → Sign up for Tiger Cloud (I have shared the link in the replies). It gives you $1,000 free credits (no card needed) → Install Tiger CLI: curl -fsSL https(:)//cli(.)tigerdata(.)com | sh → Run tiger mcp install claude-code → Give Claude Code a prompt and let it build Find the sign-up link in the replies.

Avi Chawla

14,838 Aufrufe • vor 3 Monaten

I’ve been craving a video editor with motion baked in, designed to work with my own AI agents. Claude, Codex, whatever I want to use. I don’t want to be blocked by someone else’s credit system for every little interaction. I also wanted a Figma-like canvas/editor for when I want to jump in and tweak any little detail. So we built it: Video & audio editing: cut, zoom, speed up, add B-roll. Motion, animation, 3D, custom shaders & effects, custom 3D models, dynamic components & templates, audio and beat matching. And it’s all programmable. Design your own assets directly on the canvas, bring them in from Figma or the web, or just let the agent find them, mock them up, or record what it needs automatically. You can even drop in your screen recordings and just let loose. Supercut users: yes, it’ll have first-class support for your recordings, but it’ll work with anything. I’m going to be dropping a lot more examples and tutorials, so follow along. DM me for early access if you’re willing to give feedback. We’ll open it up to everyone very soon.

I’ve been craving a video editor with motion baked in, designed to work with my own AI agents. Claude, Codex, whatever I want to use. I don’t want to be blocked by someone else’s credit system for every little interaction. I also wanted a Figma-like canvas/editor for when I want to jump in and tweak any little detail. So we built it: Video & audio editing: cut, zoom, speed up, add B-roll. Motion, animation, 3D, custom shaders & effects, custom 3D models, dynamic components & templates, audio and beat matching. And it’s all programmable. Design your own assets directly on the canvas, bring them in from Figma or the web, or just let the agent find them, mock them up, or record what it needs automatically. You can even drop in your screen recordings and just let loose. Supercut users: yes, it’ll have first-class support for your recordings, but it’ll work with anything. I’m going to be dropping a lot more examples and tutorials, so follow along. DM me for early access if you’re willing to give feedback. We’ll open it up to everyone very soon.

Neil

23,042 Aufrufe • vor 1 Monat

"Inference, if you look at it as a market, will be much, much bigger than cloud computing was pre-ChatGPT." Lightspeed’s Bucky Moore says inference is an underrated investment category in AI, and expects the market to break up into large, specialized platforms for each modality: "The GPU supply crunch that we're seeing right now is largely, as Dylan Patel has said on the show before, due to the fact that not only these consumer products, but also B2B products like Claude Code and Codex are just really taking off and creating insane demand for inference." "We're talking hundreds of billions in spend every year. And if that's true, I think there will be very, very large inference platforms built in each modality." "So there will be an inference platform for real-time video models, there will be an inference platform for open-source and custom language models, there will be an inference platform built specifically for long-running agents." "So I think we're just going to see that industry, which today looks like one industry, break up into many because of how big it is and how much room for specialization there is."

"Inference, if you look at it as a market, will be much, much bigger than cloud computing was pre-ChatGPT." Lightspeed’s Bucky Moore says inference is an underrated investment category in AI, and expects the market to break up into large, specialized platforms for each modality: "The GPU supply crunch that we're seeing right now is largely, as Dylan Patel has said on the show before, due to the fact that not only these consumer products, but also B2B products like Claude Code and Codex are just really taking off and creating insane demand for inference." "We're talking hundreds of billions in spend every year. And if that's true, I think there will be very, very large inference platforms built in each modality." "So there will be an inference platform for real-time video models, there will be an inference platform for open-source and custom language models, there will be an inference platform built specifically for long-running agents." "So I think we're just going to see that industry, which today looks like one industry, break up into many because of how big it is and how much room for specialization there is."

TBPN

33,278 Aufrufe • vor 5 Monaten

elon musk grabbed the source code openai open-sourced by accident, rewrote it in rust over a weekend, and shipped it as a free coding agent that does everything $200/mo chatgpt pro does. why pay $200 to openai and $200 to claude when this runs for $8 the swarm above is one weekend of exactly that: thousands of agents pouring through four endpoints, three paid seats billing $1.80 a task while the free fork bills $0. musk co-founded openai, walked out, and when they left codex on github under a permissive license, he forked it, stamped grok on it, and gave it away what the free version does that the $200 seat charges for: the agent · openai's own engine -> it reads your repo, writes patches, runs your tests, and loops until they pass, exactly like codex -> because under the hood it is codex, just faster and free. you are paying $200 for the paid skin of a tool now sitting on github the license · apache-2.0, un-revocable -> free to use, free to fork, free to ship inside your own product with zero strings -> openai cannot pull it back. musk made sure the license is the kind that never expires the switch · one line, no new tools -> point it at any openai-compatible or claude-compatible endpoint, including an $8 kimi backend -> same terminal, same workflow, gpt-5.6 and opus 5 just quietly lose the seat the bill · $400 down to $8 -> chatgpt pro plus claude max is $400 a month. the free agent plus an $8 kimi key does the same daily work -> that is a 98% cut, built out of openai's own source code, handed to you by the guy suing them here is the part they will fight me on: openai did not lose this to a better model, they lost it to their own license and an enemy with a weekend free. the $200 was never the tool, it was the toll, and musk just put openai's own logo on the road around it drop your $400/mo ai stack to $8. the run above is openai's own agent, rewritten free, doing the job it bills $200 a month for. the full breakdown is in the article below

starmex

111,101 Aufrufe • vor 14 Tagen

If you forced me to pick my top three matches ever, even though I had many, this is the list 👇 Number three My first main event ever. Myself vs CM Punk inside WWE Hell in a Cell for the WWE Championship. That match changed everything for me. Number two Myself and Team Hell No Daniel Bryan and Kane vs The Shield Seth Rollins Dean Ambrose and Roman Reigns in their debut match. A TLC match that was given five stars and is regarded as one of the greatest TLC matches of all time. Number one Myself vs John Cena for the WWE Championship in Chicago Illinois at Allstate Arena. I drove John through the LED Titantron board. He was stretchered out and taken away in an ambulance. I got back to my feet, but it was still ruled a tie. Honorable mention Myself vs CM Punk in another TLC match, including the TLC match we had on Raw. The majority of that match was done on the fly with very little actually planned, which makes it even more special to me. Those are my picks if you make me choose. And that’s the last bite. Feed Me More!

If you forced me to pick my top three matches ever, even though I had many, this is the list 👇 Number three My first main event ever. Myself vs CM Punk inside WWE Hell in a Cell for the WWE Championship. That match changed everything for me. Number two Myself and Team Hell No Daniel Bryan and Kane vs The Shield Seth Rollins Dean Ambrose and Roman Reigns in their debut match. A TLC match that was given five stars and is regarded as one of the greatest TLC matches of all time. Number one Myself vs John Cena for the WWE Championship in Chicago Illinois at Allstate Arena. I drove John through the LED Titantron board. He was stretchered out and taken away in an ambulance. I got back to my feet, but it was still ruled a tie. Honorable mention Myself vs CM Punk in another TLC match, including the TLC match we had on Raw. The majority of that match was done on the fly with very little actually planned, which makes it even more special to me. Those are my picks if you make me choose. And that’s the last bite. Feed Me More!

RYBACK

23,564 Aufrufe • vor 7 Monaten

I tried jack's Buzz. It's like Slack + OpenClaw + Herdr + but with some really unique features that people are sleeping on. The video below shows how it works, and some of my thoughts on the process and platform, e.g.: - Create and interact with agents on top of any harness (claude code, codex, pi, etc.) - Choose which models agents use, including local ones - Agents can delegate work and work in parallel in git worktrees - Agents are first-class citizens and work like humans (creating channels, delegating, access to chat history) - You can share AI compute within a community - It's completely open-source and decentralized Things I like: - Delegating work in chat feels natural: tag an agent, it replies in a thread with status updates as it e.g. compiles, commits, and deploys. - Shared compute: relay owners can share local compute with members, so a community could pool funds for one beefy machine running a local model and everyone uses it. - It's built on Nostr, an open protocol already tied into Bitcoin Lightning so I can imagine communities tipping each other or paying for compute/agent tasks with instant zero-fee micropayments in the future. - It ties together things like OpenClaw, an agent manager, and Slack-style chat into one tool. Things I didn't like: - You can't see what the agent is doing in a terminal. The activity view exists, but if you're used to watching a session run, this UI feels a bit abstracted. A terminal view would be great. - It feels slower than running a session in Claude Code, though no evidence to back that up. For that reason I found myself doing one-off tasks in the terminal instead. Verdict: - I really like it so far and can genuinely imagine working with a team this way. - It doesn't feel ready for big, complex tasks yet. For shallower tasks, it's perfect. - The shared compute + Nostr/Lightning angle is what really separates it from every other agent manager for me, and I think that future is coming.

I tried jack's Buzz. It's like Slack + OpenClaw + Herdr + but with some really unique features that people are sleeping on. The video below shows how it works, and some of my thoughts on the process and platform, e.g.: - Create and interact with agents on top of any harness (claude code, codex, pi, etc.) - Choose which models agents use, including local ones - Agents can delegate work and work in parallel in git worktrees - Agents are first-class citizens and work like humans (creating channels, delegating, access to chat history) - You can share AI compute within a community - It's completely open-source and decentralized Things I like: - Delegating work in chat feels natural: tag an agent, it replies in a thread with status updates as it e.g. compiles, commits, and deploys. - Shared compute: relay owners can share local compute with members, so a community could pool funds for one beefy machine running a local model and everyone uses it. - It's built on Nostr, an open protocol already tied into Bitcoin Lightning so I can imagine communities tipping each other or paying for compute/agent tasks with instant zero-fee micropayments in the future. - It ties together things like OpenClaw, an agent manager, and Slack-style chat into one tool. Things I didn't like: - You can't see what the agent is doing in a terminal. The activity view exists, but if you're used to watching a session run, this UI feels a bit abstracted. A terminal view would be great. - It feels slower than running a session in Claude Code, though no evidence to back that up. For that reason I found myself doing one-off tasks in the terminal instead. Verdict: - I really like it so far and can genuinely imagine working with a team this way. - It doesn't feel ready for big, complex tasks yet. For shallower tasks, it's perfect. - The shared compute + Nostr/Lightning angle is what really separates it from every other agent manager for me, and I think that future is coming.

Vinny

1,355,312 Aufrufe • vor 1 Monat