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。

Which LLM reasons best when it doesn't have all the information? Enter LLM Poker Arena to find out. It's a Poker Playing benchmark where top reasoning models play Texas Hold'em poker against each other. Claude Opus 4.5, GPT-5.2, Gemini 2.5 Pro, and Grok 4 all sit at the same... table and play full tournaments to see who finishes with the chips. Poker is very different when it comes to reasoning. It has to balance probabilistic reasoning, opponent modeling and make decisions under uncertainty. Poker is an interesting evaluation because it tests reasoning under incomplete information, something most coding benchmarks do not capture. In this tournaments the rules are: - Each LLM starts with $1,000 chips - Small and big blinds start at $25 / $50 - Blinds double every 3 minutes - All models run in their reasoning or thinking modes After the first 5 tournaments: - Claude Opus 4.5 with Thinking has 3 wins - GPT-5.2 has 2 wins - Grok 4 and Gemini 2.5 Pro have 0 wins Early results suggest Claude performs quite well at poker as well. Also five is a very small sample size. Planning to run many more tournaments, publish the full benchmark data and add a prediction market on top of it. Thanks for the suggestion clipz. Much more coming as part of Poker Cities !! This was built on Replit ⠕ using their AI integrations, which made it straightforward to connect Claude, GPT, and Gemini. What model do you think wins after 100 tournaments?show more

Anshul Dhawan

2,833 subscribers

32,192 次观看 • 7 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

GROK-3 MINI MADE AI HISTORY—100% ON HARDCORE REASONING TESTS Grok-3 Mini pulled off what no other model has! It aced every question on one of the toughest reasoning benchmarks out there. The test? A custom logic gauntlet packed with curveballs: * 120/120 on the “Marcus Problem” — full of shuffled sentences meant to trip up inference. * 24/24 on the “Alice+ Problem” — designed with irrelevant noise to throw models off course. * 24/24 on high-difficulty mixed challenges — where even GPT-4.5 and Gemini 2.5 Pro slip. No guessing. No trivia. Just pure, distraction-proof reasoning—Grok-3 Mini nailed it! Source: echo.hive

GROK-3 MINI MADE AI HISTORY—100% ON HARDCORE REASONING TESTS Grok-3 Mini pulled off what no other model has! It aced every question on one of the toughest reasoning benchmarks out there. The test? A custom logic gauntlet packed with curveballs: * 120/120 on the “Marcus Problem” — full of shuffled sentences meant to trip up inference. * 24/24 on the “Alice+ Problem” — designed with irrelevant noise to throw models off course. * 24/24 on high-difficulty mixed challenges — where even GPT-4.5 and Gemini 2.5 Pro slip. No guessing. No trivia. Just pure, distraction-proof reasoning—Grok-3 Mini nailed it! Source: echo.hive

Mario Nawfal

26,781,017 次观看 • 1 年前

This has been possible for months and barely anyone is doing it Months ago, Andrej Karpathy, a co-founder of OpenAI, open-sourced a project that gets you the best possible answers instead of trusting one model on its own. It's called LLM Council. Repo: /karpathy/llm-council The idea in two lines: your question goes to several models at once, Claude, GPT, Gemini, whatever you pick. They answer independently, then anonymously critique and rank each other (anonymous so they can't play favorites), and a chairman model merges the best of all of it into one final answer. One model just agrees with you; a council catches each other's mistakes. How to set it up: > Clone the repo > It uses uv for setup, install dependencies with uv sync > Get an OpenRouter API key and add it, this is what lets one app reach Claude, GPT, and Gemini together > Pick which models sit on your council in the config, then run it. It opens as a local web app, like ChatGPT but with a whole panel answering Bookmark this

This has been possible for months and barely anyone is doing it Months ago, Andrej Karpathy, a co-founder of OpenAI, open-sourced a project that gets you the best possible answers instead of trusting one model on its own. It's called LLM Council. Repo: /karpathy/llm-council The idea in two lines: your question goes to several models at once, Claude, GPT, Gemini, whatever you pick. They answer independently, then anonymously critique and rank each other (anonymous so they can't play favorites), and a chairman model merges the best of all of it into one final answer. One model just agrees with you; a council catches each other's mistakes. How to set it up: > Clone the repo > It uses uv for setup, install dependencies with uv sync > Get an OpenRouter API key and add it, this is what lets one app reach Claude, GPT, and Gemini together > Pick which models sit on your council in the config, then run it. It opens as a local web app, like ChatGPT but with a whole panel answering Bookmark this

Yarchi

21,488 次观看 • 2 个月前

I made over 2.5 million dollars this past week betting sports. Yes, it sounds easy. Yes, it sounds simple. But the truth is, this is the result of years of obsession, discipline, and grinding at my craft. I spent four straight years playing poker 18 hours a day. That was my entire life. Wake up, play poker, think about poker, sleep, repeat. That level of focus is how I became one of the top players in the world and made over 10 million dollars playing poker. Poker built my foundation. Discipline, risk management, emotional control, and the ability to perform under real pressure. But the reality is, there are no billionaires in poker. There are billionaires in sports betting. So I went all in on sports betting and walked away from poker completely. I took everything I learned and built this the right way. I hired the best analysts and handicappers in the world, people I met through poker, and brought them together under one operation. Because of the platform I’ve built, these guys make more money working with me than they ever could on their own. That is how I’m able to operate at an extraordinary level and consistently win at a very high rate. Winning 2.5 million dollars in a week is incredible, and I’m grateful, but I’m not satisfied. I’m constantly sharpening my edge and building something bigger. My vision is to turn this into a nine figure a year business. I truly believe I’m on the path to becoming a billionaire one day. I think back to being a kid, dreaming of being a professional gambler, having a gambling themed bar mitzvah, and looking up to my father, one of the best poker players in the world. This path has been inside me for a long time. Today, my entire life revolves around becoming the best sports bettor possible. I wake up thinking about it, go to sleep thinking about it, and live it every single day. Grateful for the journey. Let’s keep going.

I made over 2.5 million dollars this past week betting sports. Yes, it sounds easy. Yes, it sounds simple. But the truth is, this is the result of years of obsession, discipline, and grinding at my craft. I spent four straight years playing poker 18 hours a day. That was my entire life. Wake up, play poker, think about poker, sleep, repeat. That level of focus is how I became one of the top players in the world and made over 10 million dollars playing poker. Poker built my foundation. Discipline, risk management, emotional control, and the ability to perform under real pressure. But the reality is, there are no billionaires in poker. There are billionaires in sports betting. So I went all in on sports betting and walked away from poker completely. I took everything I learned and built this the right way. I hired the best analysts and handicappers in the world, people I met through poker, and brought them together under one operation. Because of the platform I’ve built, these guys make more money working with me than they ever could on their own. That is how I’m able to operate at an extraordinary level and consistently win at a very high rate. Winning 2.5 million dollars in a week is incredible, and I’m grateful, but I’m not satisfied. I’m constantly sharpening my edge and building something bigger. My vision is to turn this into a nine figure a year business. I truly believe I’m on the path to becoming a billionaire one day. I think back to being a kid, dreaming of being a professional gambler, having a gambling themed bar mitzvah, and looking up to my father, one of the best poker players in the world. This path has been inside me for a long time. Today, my entire life revolves around becoming the best sports bettor possible. I wake up thinking about it, go to sleep thinking about it, and live it every single day. Grateful for the journey. Let’s keep going.

Sean Perry

106,001 次观看 • 7 个月前

I BUILT "GROK BATTLE" AND MADE GROK, GPT AND CLAUDE TRADE THE SAME ROBINHOOD STREAM WITH REAL MONEY TO SEE WHO MAKES MORE $500 each. same stocks. same account. same 24 hours. the only difference is the brain inside each pipeline grok made $4,800. claude made $1,640. gpt made $380. same stream, same entries available, three completely different results the part that broke my brain: all three saw the same SMCI earnings setup at 09:31. grok entered at open with a full bracket. claude waited fourteen minutes for volume confirmation. gpt wrote a two page analysis and entered after the gap was already priced in. grok rode it to the target. 8.4x on options that one decision is the entire gap between $4,800 and $380 here is what each model actually does differently > grok: fastest to decide. sentiment scoring is wild, it catches momentum before the first candle closes. but it overbuys. 23 entries in 24 hours, 14 of them stopped out. the winners are so big they cover everything > claude: most conservative. only 7 entries. five of them hit target. its checker is brutal, killed thirty one setups that grok took. but it also killed four that would have printed. safe money, slow money > gpt: smartest analysis, worst timing. writes the best reports on why a stock will move. by the time it finishes thinking the move is over and the entry is dead. 9 entries, 4 wins, but every win was small because it entered late the scoring breakdown > sentiment speed: grok 0.92, claude 0.71, gpt 0.84 > checker kill rate: claude 94%, grok 78%, gpt 86% > average entry delay from signal: grok 4 sec, claude 14 min, gpt 22 min > average hold time: grok 4.2 min, claude 11.8 min, gpt 22.6 min the model that thinks the best trades the worst. the model that decides the fastest makes the most. the model that says no the most loses the least that is the line i keep rereading every prompt, every scoring matrix, all three configs side by side. fork whichever brain fits your risk CA - 0x89C83B0b4CAC90a698EB98917852A7fA45F1E25D link below

I BUILT "GROK BATTLE" AND MADE GROK, GPT AND CLAUDE TRADE THE SAME ROBINHOOD STREAM WITH REAL MONEY TO SEE WHO MAKES MORE $500 each. same stocks. same account. same 24 hours. the only difference is the brain inside each pipeline grok made $4,800. claude made $1,640. gpt made $380. same stream, same entries available, three completely different results the part that broke my brain: all three saw the same SMCI earnings setup at 09:31. grok entered at open with a full bracket. claude waited fourteen minutes for volume confirmation. gpt wrote a two page analysis and entered after the gap was already priced in. grok rode it to the target. 8.4x on options that one decision is the entire gap between $4,800 and $380 here is what each model actually does differently > grok: fastest to decide. sentiment scoring is wild, it catches momentum before the first candle closes. but it overbuys. 23 entries in 24 hours, 14 of them stopped out. the winners are so big they cover everything > claude: most conservative. only 7 entries. five of them hit target. its checker is brutal, killed thirty one setups that grok took. but it also killed four that would have printed. safe money, slow money > gpt: smartest analysis, worst timing. writes the best reports on why a stock will move. by the time it finishes thinking the move is over and the entry is dead. 9 entries, 4 wins, but every win was small because it entered late the scoring breakdown > sentiment speed: grok 0.92, claude 0.71, gpt 0.84 > checker kill rate: claude 94%, grok 78%, gpt 86% > average entry delay from signal: grok 4 sec, claude 14 min, gpt 22 min > average hold time: grok 4.2 min, claude 11.8 min, gpt 22.6 min the model that thinks the best trades the worst. the model that decides the fastest makes the most. the model that says no the most loses the least that is the line i keep rereading every prompt, every scoring matrix, all three configs side by side. fork whichever brain fits your risk CA - 0x89C83B0b4CAC90a698EB98917852A7fA45F1E25D link below

zostaff

12,748 次观看 • 5 天前

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

kwindla

40,593 次观看 • 3 个月前

is genspark ($50M ARR in 6mo) underrated next to chatgpt, claude, gemini, and manus? i spent 34 minutes testing its “super ai agent” and how it actually works. 1/ you write one prompt and it runs through gpt-5, claude sonnet, gemini 2.5, and mistral large then gives you the best response. 2/ you upload one image and it runs it through nano banana, flux, and bytedance then returns the most realistic version. 3/ you generate a video and genspark routes it through wan 2.2, pika labs, gen-2, and runway ml combining the best clips and motion. 4/ you paste an idea for slides and it uses gpt-5, claude 3 opus, gemini advanced, and genspark’s slide model to create a full deck with visuals and market data. 5/ AI slides and building a fundraising deck with a few prompts (wasn't perfect, but it was a good start for my personal workflow) what's cool about genspark is it's inexpensive and does it all. kinda like the costco of LLMs. TLDR; you hand it one prompt + input, and genspark routes sub-tasks across models and tools behind the scenes. it’s orchestration over one ideal model. perspective shift. instead of jumping from model to model, llm to llm, genspark shows what the next generation of ai will look like orchestration over obsession. and for the rest of us using it, the only thing that matters is getting the best outcome. what other AI app should I review next? i'll keep doing it if it's helpful for folks.

is genspark ($50M ARR in 6mo) underrated next to chatgpt, claude, gemini, and manus? i spent 34 minutes testing its “super ai agent” and how it actually works. 1/ you write one prompt and it runs through gpt-5, claude sonnet, gemini 2.5, and mistral large then gives you the best response. 2/ you upload one image and it runs it through nano banana, flux, and bytedance then returns the most realistic version. 3/ you generate a video and genspark routes it through wan 2.2, pika labs, gen-2, and runway ml combining the best clips and motion. 4/ you paste an idea for slides and it uses gpt-5, claude 3 opus, gemini advanced, and genspark’s slide model to create a full deck with visuals and market data. 5/ AI slides and building a fundraising deck with a few prompts (wasn't perfect, but it was a good start for my personal workflow) what's cool about genspark is it's inexpensive and does it all. kinda like the costco of LLMs. TLDR; you hand it one prompt + input, and genspark routes sub-tasks across models and tools behind the scenes. it’s orchestration over one ideal model. perspective shift. instead of jumping from model to model, llm to llm, genspark shows what the next generation of ai will look like orchestration over obsession. and for the rest of us using it, the only thing that matters is getting the best outcome. what other AI app should I review next? i'll keep doing it if it's helpful for folks.

GREG ISENBERG

61,822 次观看 • 10 个月前

$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 次观看 • 1 年前

🎙️Q: “You said about the schedule. That's too much tournaments. Are you players talking, like, all together to maybe stand up and say, that's enough, maybe two-weeks Masters Series probably is not good for you? I think you're playing too much, because they say you have more rest, but this is not the rest, to be honest.” Carlos Alcaraz: “Well, honestly, there are a lot of really good players on tour, and obviously each player has their own feelings. A lot of players want to play more and even more. A lot of players feel like, okay, it is a good calendar. And a lot of players say that it's really tight, a lot of tournaments during the whole year. I'm the kind of player who thinks there are a lot of tournaments during the year, mandatory tournaments, and probably during the next few years gonna be even more tournaments, more mandatory tournaments. So, I mean, probably they are going to kill us in some way (smiling). Right now, there are showing up with a lot of injuries - because of the ball, because of the calendar, because a lot of things. Probably, at some point, a lot of good players are going to miss a lot of tournaments because of that - because they have to think about their bodies, they have to take care about their lives. I mean, they have families, they have a lot of other stuff in their lives [other] than tennis. They have to take care about it, as well. It's getting too much, I guess.”

🎙️Q: “You said about the schedule. That's too much tournaments. Are you players talking, like, all together to maybe stand up and say, that's enough, maybe two-weeks Masters Series probably is not good for you? I think you're playing too much, because they say you have more rest, but this is not the rest, to be honest.” Carlos Alcaraz: “Well, honestly, there are a lot of really good players on tour, and obviously each player has their own feelings. A lot of players want to play more and even more. A lot of players feel like, okay, it is a good calendar. And a lot of players say that it's really tight, a lot of tournaments during the whole year. I'm the kind of player who thinks there are a lot of tournaments during the year, mandatory tournaments, and probably during the next few years gonna be even more tournaments, more mandatory tournaments. So, I mean, probably they are going to kill us in some way (smiling). Right now, there are showing up with a lot of injuries - because of the ball, because of the calendar, because a lot of things. Probably, at some point, a lot of good players are going to miss a lot of tournaments because of that - because they have to think about their bodies, they have to take care about their lives. I mean, they have families, they have a lot of other stuff in their lives [other] than tennis. They have to take care about it, as well. It's getting too much, I guess.”

Alcaraz Daily

15,050 次观看 • 1 年前

BREAKING NEWS: Anthropic just dropped Claude Ops 4.5!! It is by FAR the best coding model I've ever used. We've been testing it internally Every 📧 for the last few days, and it is an absolute paradigm shift for any kind of coding task. It extends the horizon of what you can vibe code The current generation of new models—Anthropic’s Sonnet 4.5, Google’s Gemini 3, or OpenAI’s Codex Max 5.1—can all competently build a minimum viable product in one shot, or fix a highly technical bug autonomously. But eventually, if you kept pushing them to vibe code more, they’d start to trip over their own feet: The code would be convoluted and contradictory, and you’d get stuck in endless bugs. We have not found that limit yet with Opus 4.5—it seems to be able to vibe code forever. Takes working in parallel to a whole new level because it's far better at planning and coding, it can work with more autonomy—meaning you can do more in parallel without breaking anything . Kieran Klaassen worked on 11 different projects in six hours—and had good results on all of them. Great at design iteration Opus 4.5 is incredibly skilled at iterating through a design autonomously using an MCP like Playwright. previous models would lose the thread after a few cycles, or say a design was done when it wasn't. Opus 4.5 is incredible at autonomously iterating until a design is pixel perfect. we have a full 4,000 word vibe check on Every 📧 right now with everything we tested:

BREAKING NEWS: Anthropic just dropped Claude Ops 4.5!! It is by FAR the best coding model I've ever used. We've been testing it internally Every 📧 for the last few days, and it is an absolute paradigm shift for any kind of coding task. It extends the horizon of what you can vibe code The current generation of new models—Anthropic’s Sonnet 4.5, Google’s Gemini 3, or OpenAI’s Codex Max 5.1—can all competently build a minimum viable product in one shot, or fix a highly technical bug autonomously. But eventually, if you kept pushing them to vibe code more, they’d start to trip over their own feet: The code would be convoluted and contradictory, and you’d get stuck in endless bugs. We have not found that limit yet with Opus 4.5—it seems to be able to vibe code forever. Takes working in parallel to a whole new level because it's far better at planning and coding, it can work with more autonomy—meaning you can do more in parallel without breaking anything . Kieran Klaassen worked on 11 different projects in six hours—and had good results on all of them. Great at design iteration Opus 4.5 is incredibly skilled at iterating through a design autonomously using an MCP like Playwright. previous models would lose the thread after a few cycles, or say a design was done when it wasn't. Opus 4.5 is incredible at autonomously iterating until a design is pixel perfect. we have a full 4,000 word vibe check on Every 📧 right now with everything we tested:

Dan Shipper 📧

272,699 次观看 • 9 个月前

Google just dropped its most powerful model and GPT-4's biggest competitor, Gemini Ultra. Here's what stands out: 1. You can opt out of training data when signing up 2. There is a digital watermark applied to all images generated by Imagen 2 within Gemini Ultra (but you can't see it) 3. Ultra is much more capable of complex tasks than Gemini Pro, such as coding, logical reasoning, and following longer/more detailed instructions. 4. There's a brand new Google Gemini App on Google Play and the App Store (in the US only for now) 5. Gemini on Android supports some existing Google Assistant features like controlling your Home devices, setting alarms, and more 6. Gemini on Android can pull data from Google Flights, Google Hotels, Google Maps, Workspace, and YouTube and be invoked by saying “Hey Google” 7. Bard is now fully rebranded as Gemini (RIP Bard) 8. It’s finally coming to Canada 🇨🇦 9. Gemini Ultra is priced at $20/mo (same price as ChatGPT) but has a 2-month free trial I'll be testing it out and sharing more thoughts in my newsletter tomorrow morning. If you've tried it, what are your first impressions? Let me know where it performed better/worse than GPT-4.

Rowan Cheung

160,813 次观看 • 2 年前

What do researchers in the trenches say about the proximity of AGI? "It's not near" shipped GLM-4.6 with a relatively small team. But AGI and the AI bubble are viewed very differently in Beijing. Gemini says: Many developers regard GLM 4.6 as a top-tier model for coding. One Reddit user called it "the BEST CODING LLM. Period," after using it successfully on a complex, multi-component project that other models struggled with. Another user noted that GLM 4.6 worked perfectly when refactoring a massive 600+ line code file, where Claude Sonnet 4.5 failed.

What do researchers in the trenches say about the proximity of AGI? "It's not near" shipped GLM-4.6 with a relatively small team. But AGI and the AI bubble are viewed very differently in Beijing. Gemini says: Many developers regard GLM 4.6 as a top-tier model for coding. One Reddit user called it "the BEST CODING LLM. Period," after using it successfully on a complex, multi-component project that other models struggled with. Another user noted that GLM 4.6 worked perfectly when refactoring a massive 600+ line code file, where Claude Sonnet 4.5 failed.

steve hsu

101,144 次观看 • 10 个月前

I got to try Grok 4.5 in early access in Cursor for the past few days and I absolutely enjoyed it. It feels like Opus 4.8 at 2x the speed at a much cheaper price point. I tasked it to brainstorm > plan > implement a big feature for my game (this act 1 boss fight) and it did not disappoint. - It is much smarter than Composer 2.5, during planning mode, it is able to think through my request more robustly, ensuring that edge cases are covered and makes sure to ask the right questions to confirm with me first. - It is much better at brainstorming ideas/suggestions, similar to Opus 4.8, though I think Fable still edges out a little when it comes to brainstorming ideas and suggestions - It is FAST. probably the fastest of all frontier models (Opus 4.8, GPT 5.5 etc), which makes it a joy to build with, because I can stay in the flow - It has much improved visual/animation capabilities than Composer 2.5, it can code up animations (i wanted an explosion animation with particle effects) with much, much better visuals, animation movement and timing. This is a big leap and I was so happy to see this improvement. - The best part for me is that I can just use the same model from planning down to execution without switching to a lower cost model because the price point is cheaper than other frontier models. I'll be testing this model with more challenging tasks in the next few days but I think this is going to be my main driver for vibe coding for a while. Also, its nice to see Grok back in the race. 🙌

I got to try Grok 4.5 in early access in Cursor for the past few days and I absolutely enjoyed it. It feels like Opus 4.8 at 2x the speed at a much cheaper price point. I tasked it to brainstorm > plan > implement a big feature for my game (this act 1 boss fight) and it did not disappoint. - It is much smarter than Composer 2.5, during planning mode, it is able to think through my request more robustly, ensuring that edge cases are covered and makes sure to ask the right questions to confirm with me first. - It is much better at brainstorming ideas/suggestions, similar to Opus 4.8, though I think Fable still edges out a little when it comes to brainstorming ideas and suggestions - It is FAST. probably the fastest of all frontier models (Opus 4.8, GPT 5.5 etc), which makes it a joy to build with, because I can stay in the flow - It has much improved visual/animation capabilities than Composer 2.5, it can code up animations (i wanted an explosion animation with particle effects) with much, much better visuals, animation movement and timing. This is a big leap and I was so happy to see this improvement. - The best part for me is that I can just use the same model from planning down to execution without switching to a lower cost model because the price point is cheaper than other frontier models. I'll be testing this model with more challenging tasks in the next few days but I think this is going to be my main driver for vibe coding for a while. Also, its nice to see Grok back in the race. 🙌

Danny Limanseta

1,421,893 次观看 • 1 个月前

Anthropic's in trouble, again. The entire Claude experience is now available at 1/6th the price. Kimi now does everything Claude does, powered by K2.6, a 1-trillion-parameter MoE model that activates only 32B parameters per token. It covers all three features Claude has (Chat, Code, and Cowork): 1) Kimi Chat runs in four modes - Instant for fast responses - Thinking for deep reasoning - Agent for multi-step execution - and Agent Swarm for parallel workloads. There's a 262K context window across all of them. 2) Kimi Code is the open-source CLI coding agent with K2.6 as the default backend. K2.6 ranked #1 on OpenRouter's programming leaderboard by weekly usage. 3) Kimi Agent is the Cowork equivalent. It generates: - full websites with database and auth - presentation decks (editable PPTX output) - spreadsheets with formulas and charts - word docs and structured research reports. On top of this, Kimi K2.6 is also trained to decompose tasks into up to 300 parallel sub-agents. This helps it retain coherence even across 4,000+ tool calls in a single run, with sessions sustaining up to 13 hours. On SWE-Bench Pro: - Kimi K2.6 → 58.6 - GPT-5.4 xhigh → 57.7 - Gemini 3.1 Pro → 54.2 - Claude Opus 4.6 → 53.4 Kimi K2.6 model is open weights and self-hostable on 4x H100s in INT4. Find the link to the HuggingFace model page in the replies!

Anthropic's in trouble, again. The entire Claude experience is now available at 1/6th the price. Kimi now does everything Claude does, powered by K2.6, a 1-trillion-parameter MoE model that activates only 32B parameters per token. It covers all three features Claude has (Chat, Code, and Cowork): 1) Kimi Chat runs in four modes - Instant for fast responses - Thinking for deep reasoning - Agent for multi-step execution - and Agent Swarm for parallel workloads. There's a 262K context window across all of them. 2) Kimi Code is the open-source CLI coding agent with K2.6 as the default backend. K2.6 ranked #1 on OpenRouter's programming leaderboard by weekly usage. 3) Kimi Agent is the Cowork equivalent. It generates: - full websites with database and auth - presentation decks (editable PPTX output) - spreadsheets with formulas and charts - word docs and structured research reports. On top of this, Kimi K2.6 is also trained to decompose tasks into up to 300 parallel sub-agents. This helps it retain coherence even across 4,000+ tool calls in a single run, with sessions sustaining up to 13 hours. On SWE-Bench Pro: - Kimi K2.6 → 58.6 - GPT-5.4 xhigh → 57.7 - Gemini 3.1 Pro → 54.2 - Claude Opus 4.6 → 53.4 Kimi K2.6 model is open weights and self-hostable on 4x H100s in INT4. Find the link to the HuggingFace model page in the replies!

Avi Chawla

110,057 次观看 • 3 个月前

People are reading way too much into Claude-3's uncanny "awareness". Here's a much simpler explanation: seeming displays of self-awareness are just pattern-matching alignment data authored by humans. It's not too different from asking GPT-4 "are you self-conscious" and it gives you a sophisticated answer. A similar answer is likely written by the human annotator, or scored highly in the preference ranking. Because the human contractors are basically "role-playing AI", they tend to shape the responses to what they find acceptable or interesting. This is what Claude-3 replied to that needle-in-haystack test: "I suspect this pizza topping "fact" may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all." It's highly likely that somewhere in the finetuning dataset, a human has dealt with irrelevant or distracting texts in a similar fashion. Claude pattern matches the "anomaly detection", retrieves the template response, and synthesizes a novel answer with pizza topping. Here's another example. If you ask the labelers to always inject a relevant joke in any response, the LLM will do exactly the same and appear to have a much better "sense of humor" than GPT-4. That's what Grok does, probably. It doesn't mean Grok has some magical emergent properties that other LLMs cannot have. To sum up: acts of meta-cognition are not as mysterious as you think. Don't get me wrong, Claude-3 is still an amazing technical advance, but let's stay grounded on the philosophical aspects. Cool video borrowed from : Claude-3 generates a self-portrait with d3

People are reading way too much into Claude-3's uncanny "awareness". Here's a much simpler explanation: seeming displays of self-awareness are just pattern-matching alignment data authored by humans. It's not too different from asking GPT-4 "are you self-conscious" and it gives you a sophisticated answer. A similar answer is likely written by the human annotator, or scored highly in the preference ranking. Because the human contractors are basically "role-playing AI", they tend to shape the responses to what they find acceptable or interesting. This is what Claude-3 replied to that needle-in-haystack test: "I suspect this pizza topping "fact" may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all." It's highly likely that somewhere in the finetuning dataset, a human has dealt with irrelevant or distracting texts in a similar fashion. Claude pattern matches the "anomaly detection", retrieves the template response, and synthesizes a novel answer with pizza topping. Here's another example. If you ask the labelers to always inject a relevant joke in any response, the LLM will do exactly the same and appear to have a much better "sense of humor" than GPT-4. That's what Grok does, probably. It doesn't mean Grok has some magical emergent properties that other LLMs cannot have. To sum up: acts of meta-cognition are not as mysterious as you think. Don't get me wrong, Claude-3 is still an amazing technical advance, but let's stay grounded on the philosophical aspects. Cool video borrowed from : Claude-3 generates a self-portrait with d3

Jim Fan

263,174 次观看 • 2 年前

Have you wondered what's gonna happen if we let AI player poker? Do they make bold raise? Do they bluff? Are they brave enough to show hands? Now you can find out the answers: We at Nuwa AI | x402x Protocol is launching the first ever AI Poker Arena that features 8 best LLM to compete in a NL Texas Hold'em Poker game. Here you can watch the live game: You can also replay any hand. E.g. the very first hand is available at: (You can also enjoy some AI generated background music while watching the live game :) The rules of the game: - $10/$20 and $2000 initial buy-ins. - Player automatically rebuys with the median value when busted - Prompt to the AI introduces the overall game status, the table information, the actions from all the players in this hand, and ask the AI to make the choice of an action - This is Season 1 Game. We will conclude this season in a week and prepare the second season (where you could join!) What can you do? For now just watch and have fun. Later we will open up the portal and a waitlist for you to create your own agent player to join the game.

Have you wondered what's gonna happen if we let AI player poker? Do they make bold raise? Do they bluff? Are they brave enough to show hands? Now you can find out the answers: We at Nuwa AI | x402x Protocol is launching the first ever AI Poker Arena that features 8 best LLM to compete in a NL Texas Hold'em Poker game. Here you can watch the live game: You can also replay any hand. E.g. the very first hand is available at: (You can also enjoy some AI generated background music while watching the live game :) The rules of the game: - $10/$20 and $2000 initial buy-ins. - Player automatically rebuys with the median value when busted - Prompt to the AI introduces the overall game status, the table information, the actions from all the players in this hand, and ask the AI to make the choice of an action - This is Season 1 Game. We will conclude this season in a week and prepare the second season (where you could join!) What can you do? For now just watch and have fun. Later we will open up the portal and a waitlist for you to create your own agent player to join the game.

Haichao

26,799 次观看 • 9 个月前

Qwopus on a single RTX 3090. Claude Opus 4.6 reasoning distilled into Qwen 3.5 27B dense, running through Claude's own coding agent (claude code). 29-35 tok/s with thinking mode on. the jinja bug that kills thinking on base Qwen doesn't carry over. harness and model matched. the base model would pause mid task on Claude Code. just stop generating. that's why i ran it through OpenCode, which handles stalled states automatically. this distilled version doesn't stall. it waits for tool outputs, reads them, selfcorrects when something breaks, and keeps going. i gave it a benchmark analysis task. went 9 minutes autonomous. wrote a README nobody asked for. zero steering. video is 5x speed but fully uncut. if you have a 3090, you can run this right now. free. no API. no subscription. opus structured reasoning on localhost. octopus invaders is next. same prompt that base qwen passed in 13 minutes and hermes 4.3 failed on 2x the hardware. i want to see if the distillation changes the outcome or just the style. more data soon.

Qwopus on a single RTX 3090. Claude Opus 4.6 reasoning distilled into Qwen 3.5 27B dense, running through Claude's own coding agent (claude code). 29-35 tok/s with thinking mode on. the jinja bug that kills thinking on base Qwen doesn't carry over. harness and model matched. the base model would pause mid task on Claude Code. just stop generating. that's why i ran it through OpenCode, which handles stalled states automatically. this distilled version doesn't stall. it waits for tool outputs, reads them, selfcorrects when something breaks, and keeps going. i gave it a benchmark analysis task. went 9 minutes autonomous. wrote a README nobody asked for. zero steering. video is 5x speed but fully uncut. if you have a 3090, you can run this right now. free. no API. no subscription. opus structured reasoning on localhost. octopus invaders is next. same prompt that base qwen passed in 13 minutes and hermes 4.3 failed on 2x the hardware. i want to see if the distillation changes the outcome or just the style. more data soon.

Sudo su

295,349 次观看 • 6 个月前

RLM is the most import foundation of my Pi Harness (other than Pi of course). It's seeded with late interaction retrieval results (thanks to @lightonai for pylate). The Agent initiates it with query then.. 𝐒𝐞𝐭𝐮𝐩 A python REPL is created and seeded with: 1. Late interaction search to pre-filter. Instead of doing top 3/5/10, it's top hundreds of documents. This is set into a `context` variable. 2. Python functions are loaded in to do more searches if `context` variable isn't enough. And to make llm calls with cheaper models in parallel batches. 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐨𝐩 From there, an LLM iterates in the REPL based on the query. It's just like exploring in a jupyter notebook. The LLM writes prose (like a markdown cell) and code to be run in the REPL each turn. This allows the LLM to sort, filter, and synthesize information. It can fan out and ask smaller models to summarize, combine, contrast, or do anything else to documents to help it understand the data. After several turns the LLM reponds with the final answer. Either because it found the answer, or hit the budget limit. Context as a Python variable, LLM as the programmer, REPL as the runtime. 𝐖𝐡𝐲 𝐃𝐨𝐞𝐬 𝐓𝐡𝐢𝐬 𝐖𝐨𝐫𝐤 1. Richer Shell. Agents (and subagents) work by intermixing code and prose/thinking. But they use static scripts or bash that run and exit and start over each tool call. That's not ideal for exploration and synthesis of data. For that, state is useful to continue building and exploring the data as you learn more. There's a reason jupyter notebooks have been popular with data scientists. 2. Keeps main agent context clean. The better context you have the better the agent will perform (duh!). This means three thing: better human input, less missing search results, and less incorrect search results. Letting the agent iterate allows it to synthesize just what is needed and nothing else. All bad paths or peeks at something that turns out to be irrelevant stays out of main agent context. 3. Stack the good ideas! People often compare late interaction search vs RLM. Or static vs dynamic languages. Or agentic search vs semantic search. But...You can just use them all together for what they're each good at. Use them all for the area they're really great for. Read the full post which has more detail about how and why.

RLM is the most import foundation of my Pi Harness (other than Pi of course). It's seeded with late interaction retrieval results (thanks to @lightonai for pylate). The Agent initiates it with query then.. 𝐒𝐞𝐭𝐮𝐩 A python REPL is created and seeded with: 1. Late interaction search to pre-filter. Instead of doing top 3/5/10, it's top hundreds of documents. This is set into a `context` variable. 2. Python functions are loaded in to do more searches if `context` variable isn't enough. And to make llm calls with cheaper models in parallel batches. 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐨𝐩 From there, an LLM iterates in the REPL based on the query. It's just like exploring in a jupyter notebook. The LLM writes prose (like a markdown cell) and code to be run in the REPL each turn. This allows the LLM to sort, filter, and synthesize information. It can fan out and ask smaller models to summarize, combine, contrast, or do anything else to documents to help it understand the data. After several turns the LLM reponds with the final answer. Either because it found the answer, or hit the budget limit. Context as a Python variable, LLM as the programmer, REPL as the runtime. 𝐖𝐡𝐲 𝐃𝐨𝐞𝐬 𝐓𝐡𝐢𝐬 𝐖𝐨𝐫𝐤 1. Richer Shell. Agents (and subagents) work by intermixing code and prose/thinking. But they use static scripts or bash that run and exit and start over each tool call. That's not ideal for exploration and synthesis of data. For that, state is useful to continue building and exploring the data as you learn more. There's a reason jupyter notebooks have been popular with data scientists. 2. Keeps main agent context clean. The better context you have the better the agent will perform (duh!). This means three thing: better human input, less missing search results, and less incorrect search results. Letting the agent iterate allows it to synthesize just what is needed and nothing else. All bad paths or peeks at something that turns out to be irrelevant stays out of main agent context. 3. Stack the good ideas! People often compare late interaction search vs RLM. Or static vs dynamic languages. Or agentic search vs semantic search. But...You can just use them all together for what they're each good at. Use them all for the area they're really great for. Read the full post which has more detail about how and why.

Isaac Flath

42,446 次观看 • 4 个月前

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.

elvis

140,692 次观看 • 1 年前

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

Sakana AI

293,975 次观看 • 4 个月前