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。

Fable 5 on Hyperagent is producing the most creative, ambitious work we've ever seen from our agents. They're self-improving for hours towards open-ended goals. Visual reasoning has spiked noticeably. Outputs are consistently higher quality than Opus, occasionally at lower cost. 5 of our test cases below vs. Opus 4.8... show more

Hyperagent

9,649 subscribers

1,223,628 次观看 • 1 个月前 •via X (Twitter)

教育新闻政治科学技术

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

Anthropic Head of Product: "Fable 5 is the best model we've ever shipped for agentic work. Our engineers now ship 8x more code than they did in past years." in 42 minutes, the Anthropic team shows how to build self-improving agent systems with Fable 5: managed agents, dreaming, dynamic workflows worth more than a $500 agent building course live from Anthropic's latest stage in Japan, just hours ago i turned everything they showed on stage into a step-by-step guide on building your own agent you can find the full guide below

Anthropic Head of Product: "Fable 5 is the best model we've ever shipped for agentic work. Our engineers now ship 8x more code than they did in past years." in 42 minutes, the Anthropic team shows how to build self-improving agent systems with Fable 5: managed agents, dreaming, dynamic workflows worth more than a $500 agent building course live from Anthropic's latest stage in Japan, just hours ago i turned everything they showed on stage into a step-by-step guide on building your own agent you can find the full guide below

Anatoli Kopadze

79,808 次观看 • 1 个月前

I couldn't believe the results of this benchmarking test. I blind-tested 6 frontier models — Sonnet 5, Opus 4.8, Fable 5, GPT 5.5, GLM 5.2, and Kimi 2.5 — on knowledge work, not engineering. Writing. Analysis. Design. Vibe coding. Subagent simulations. The stuff we actually do all day. Same prompts, each model running its own subagent, every output graded blind before the labels were revealed. The results: → GLM 5.2 (the dirt-cheap open-source model from China) took first → Sonnet 5 took second, beating Fable 5 for knowledge work → Opus 4.8 got whooped Then I had Composer 2.5 judge every output independently — and it picked Fable 5 almost across the board. My blind picks and the AI judge completely disagreed. Full breakdown, method, and prompts in the video below. Which model are you running as your daily driver right now?

I couldn't believe the results of this benchmarking test. I blind-tested 6 frontier models — Sonnet 5, Opus 4.8, Fable 5, GPT 5.5, GLM 5.2, and Kimi 2.5 — on knowledge work, not engineering. Writing. Analysis. Design. Vibe coding. Subagent simulations. The stuff we actually do all day. Same prompts, each model running its own subagent, every output graded blind before the labels were revealed. The results: → GLM 5.2 (the dirt-cheap open-source model from China) took first → Sonnet 5 took second, beating Fable 5 for knowledge work → Opus 4.8 got whooped Then I had Composer 2.5 judge every output independently — and it picked Fable 5 almost across the board. My blind picks and the AI judge completely disagreed. Full breakdown, method, and prompts in the video below. Which model are you running as your daily driver right now?

JJ Englert

10,737 次观看 • 15 天前

Claude Fable 5 makes new things possible. I one shot a full horror game with it. One prompt. I have never seen AI output a game like this. This is the leap I've been waiting for. 80.3 on SWE-Bench Pro vs Opus 4.8 at 69.2, the biggest jump I have ever seen on that benchmark. It passed GPT 5.5 in both coding and intelligence on Artificial Analysis. I already canceled my $200 ChatGPT Pro plan. It fixed a Bridge Voice bug in one shot that GPT 5.5, Opus 4.7, and Opus 4.8 all failed to crack for 2 months. 10/10. Best model I have ever used. Full review below.

Claude Fable 5 makes new things possible. I one shot a full horror game with it. One prompt. I have never seen AI output a game like this. This is the leap I've been waiting for. 80.3 on SWE-Bench Pro vs Opus 4.8 at 69.2, the biggest jump I have ever seen on that benchmark. It passed GPT 5.5 in both coding and intelligence on Artificial Analysis. I already canceled my $200 ChatGPT Pro plan. It fixed a Bridge Voice bug in one shot that GPT 5.5, Opus 4.7, and Opus 4.8 all failed to crack for 2 months. 10/10. Best model I have ever used. Full review below.

BridgeMind

19,020 次观看 • 1 个月前

A security guard turned $900 into $187,000 on Polymarket. Guarded empty buildings for 10 years at $16/hour. One sleepless shift he built a BTC bot with Claude Opus 4.8. Never wrote a single line of code. I found his wallet. Reverse-engineered the entire system. Here is what he actually did. Gave Claude new Fable 5 two prompts at the start of his shift. Opus did not write the bot. It drew the blueprint: data, signals, backtest, risk, execution. Then handed that blueprint to a swarm of 300 agents. How it works: → Decompose: Claude Fable 5 breaks the goal into a task tree → Dispatch: 300 agents fan out (some pulling Binance data, some writing code, some backtesting 5 years of data) → Execute: 4,000 steps in one 14-hour run → Review: Claude Fable 5 reads it all back and kills whatever drifted The brain never wrote a line. The hands never made a call. By sunrise he had what a 6-engineer quant team ships in a quarter. A live BTC bot on Polymarket. Backtested. Risk-capped. Running 24/7. For 10 years he watched empty buildings. Now 300 agents watch the market while he sleeps. I rebuilt this exact system with Claude new Fable 5 You only need Claude + device + 1 hour to deploy. Giving this free for 24 hours. To get it: 1. Comment "Claude" 2. Like and retweet this 3. Follow me Himanshu Kumar so I can DM you

A security guard turned $900 into $187,000 on Polymarket. Guarded empty buildings for 10 years at $16/hour. One sleepless shift he built a BTC bot with Claude Opus 4.8. Never wrote a single line of code. I found his wallet. Reverse-engineered the entire system. Here is what he actually did. Gave Claude new Fable 5 two prompts at the start of his shift. Opus did not write the bot. It drew the blueprint: data, signals, backtest, risk, execution. Then handed that blueprint to a swarm of 300 agents. How it works: → Decompose: Claude Fable 5 breaks the goal into a task tree → Dispatch: 300 agents fan out (some pulling Binance data, some writing code, some backtesting 5 years of data) → Execute: 4,000 steps in one 14-hour run → Review: Claude Fable 5 reads it all back and kills whatever drifted The brain never wrote a line. The hands never made a call. By sunrise he had what a 6-engineer quant team ships in a quarter. A live BTC bot on Polymarket. Backtested. Risk-capped. Running 24/7. For 10 years he watched empty buildings. Now 300 agents watch the market while he sleeps. I rebuilt this exact system with Claude new Fable 5 You only need Claude + device + 1 hour to deploy. Giving this free for 24 hours. To get it: 1. Comment "Claude" 2. Like and retweet this 3. Follow me Himanshu Kumar so I can DM you

Himanshu Kumar

54,480 次观看 • 1 个月前

We’re open sourcing the first document OCR benchmark for the agentic era, ParseBench. Document parsing is the foundation of every AI agent that works with real-world files. ParseBench is a benchmark that measures parsing quality specifically for agent knowledge work: ✅ It optimizes for semantic correctness (instead of exact similarity) ✅ It has the most comprehensive distribution of real-world enterprise documents It contains ~2,000 human-verified enterprise document pages with 167,000+ test rules across five dimensions that matter most: tables, charts, content faithfulness, semantic formatting, and visual grounding. We benchmarked 14 known document parsers on ParseBench, from frontier/OSS VLMs to specialized parsers to LlamaParse. Here are some of our findings: 💡 Increasing compute budget yields diminishing returns - Gemini/gpt-5-mini/haiku gain 3-5 points from minimal to high thinking, at 4x the cost. 💡 Charts are the most polarizing dimension for evaluation. Most specialized parsers score below 6%, while some VLM-based parsers do a bit better. 💡 VLMs are great at visual understanding but terrible at layout extraction. GPT-5-mini/haiku score below 10% on our visual grounding task, all specialized parsers do much better. 💡 No method crushes all 5 dimensions at once, but LlamaParse achieves the highest overall score at 84.9%, and is the leader in 4 out of the 5 dimensions. This is by far the deepest technical work that we’ve published as a company. I would encourage you to start with our blog and explore our links to Hugging Face to GitHub. All the details are in our full 35-page (!!) ArXiv whitepaper. 🌐: Blog: 📄 Paper: 💻 Code: 📊 Dataset: 🎥 YouTube:

We’re open sourcing the first document OCR benchmark for the agentic era, ParseBench. Document parsing is the foundation of every AI agent that works with real-world files. ParseBench is a benchmark that measures parsing quality specifically for agent knowledge work: ✅ It optimizes for semantic correctness (instead of exact similarity) ✅ It has the most comprehensive distribution of real-world enterprise documents It contains ~2,000 human-verified enterprise document pages with 167,000+ test rules across five dimensions that matter most: tables, charts, content faithfulness, semantic formatting, and visual grounding. We benchmarked 14 known document parsers on ParseBench, from frontier/OSS VLMs to specialized parsers to LlamaParse. Here are some of our findings: 💡 Increasing compute budget yields diminishing returns - Gemini/gpt-5-mini/haiku gain 3-5 points from minimal to high thinking, at 4x the cost. 💡 Charts are the most polarizing dimension for evaluation. Most specialized parsers score below 6%, while some VLM-based parsers do a bit better. 💡 VLMs are great at visual understanding but terrible at layout extraction. GPT-5-mini/haiku score below 10% on our visual grounding task, all specialized parsers do much better. 💡 No method crushes all 5 dimensions at once, but LlamaParse achieves the highest overall score at 84.9%, and is the leader in 4 out of the 5 dimensions. This is by far the deepest technical work that we’ve published as a company. I would encourage you to start with our blog and explore our links to Hugging Face to GitHub. All the details are in our full 35-page (!!) ArXiv whitepaper. 🌐: Blog: 📄 Paper: 💻 Code: 📊 Dataset: 🎥 YouTube:

Jerry Liu

107,866 次观看 • 3 个月前

It’s #SunDay! Here’s your space weather report for the week of June 5 - 11: • 1 M-class flare • 5 C-class flares • 36 coronal mass ejections • 1 geomagnetic storm (as of June 12, though another one is currently forecast for June 13) This video from NASA’s Solar Dynamics Observatory shows the week’s activity. This week’s most powerful flare was an M1.8 on June 6, which erupts from the southeastern (lower left) region of the Sun’s disk at about 0:42 in this video. That eruption also produced a coronal mass ejection that reached Earth three days later, but the bulk of the blast was oriented to the south and east of Earth, so the effect on our planet was minimal. Learn more about solar flares and coronal mass ejections:

It’s #SunDay! Here’s your space weather report for the week of June 5 - 11: • 1 M-class flare • 5 C-class flares • 36 coronal mass ejections • 1 geomagnetic storm (as of June 12, though another one is currently forecast for June 13) This video from NASA’s Solar Dynamics Observatory shows the week’s activity. This week’s most powerful flare was an M1.8 on June 6, which erupts from the southeastern (lower left) region of the Sun’s disk at about 0:42 in this video. That eruption also produced a coronal mass ejection that reached Earth three days later, but the bulk of the blast was oriented to the south and east of Earth, so the effect on our planet was minimal. Learn more about solar flares and coronal mass ejections:

NASA Space Alerts

47,540 次观看 • 1 个月前

We’re launching a coin for a new vertical within the world of intellectual property, focused entirely on AI data licensing. AI companies are spending billions of dollars on data (video, text, sound, etc) to train their models. Kled is the middleman. We partner with top film studios and production houses, label their video content, and sell it to AI companies like OpenAI to help train models like ChatGPT, Sora, Midjourney and more. So far, we've already sold millions of hours of data. Our next drop includes 3 million hours of footage, with deals ranging from $2–5 per minute. We've strategically raised $2 million from investors such as K5 Global, who back Uber, Xai and more, alongside creators like Alex Botez and Panda Boi who stand by our platform and its goals. We're excited to bring this project of ours on chain, where we find that it'll benefit most from the space's interactivity and leaders alike. $KLED is now live on Believe

We’re launching a coin for a new vertical within the world of intellectual property, focused entirely on AI data licensing. AI companies are spending billions of dollars on data (video, text, sound, etc) to train their models. Kled is the middleman. We partner with top film studios and production houses, label their video content, and sell it to AI companies like OpenAI to help train models like ChatGPT, Sora, Midjourney and more. So far, we've already sold millions of hours of data. Our next drop includes 3 million hours of footage, with deals ranging from $2–5 per minute. We've strategically raised $2 million from investors such as K5 Global, who back Uber, Xai and more, alongside creators like Alex Botez and Panda Boi who stand by our platform and its goals. We're excited to bring this project of ours on chain, where we find that it'll benefit most from the space's interactivity and leaders alike. $KLED is now live on Believe

Avi Patel

476,361 次观看 • 1 年前

GPT-5.6 vs GPT-5.5 on my custom spaceship prompt. I gave both models the exact same custom prompt. This is also the same prompt I previously gave to Fable 5. For context, GPT-5.6 Pro worked for 87 minutes, while GPT-5.5 Extra High worked for 34 minutes and 42 seconds. As I’ve said before, based on great authority GPT-5.6 will be an incremental/soldi improvement over GPT-5.5, not a “Fable killer.” My rough expectation has been that it would trade blows with Fable 5 on some benchmarks, maybe win around half depending on the category, but not clearly surpass it overall. And again fable five will have bigger model smell, but this was expected. After testing this coding output, that view feels pretty accurate. GPT-5.6 is clearly better than GPT-5.5 in several visual areas. The lighting, shading, chairs, object details, and exterior of the spaceship looked noticeably stronger. The scene was also easier to test. I do want to give GPT-5.5 credit though. It built out the rooms much much better and the planets looked better than GPT-5.6’s. It was also interesting that both GPT-5.5 and GPT-5.6 produced better-looking planets than Fable 5 in this specific test. The downside with GPT-5.5 was stability. The game was much glitchier and harder to test compared to GPT-5.6. But when it comes to the core of the demo, which is the spaceship itself, Fable 5 still beat both models pretty comfortably. GPT-5.6 is impressive, but from this test, it looks exactly like what I expected which was a meaningful incremental improvement over GPT-5.5, at least for indie game demos, but not something that replaces Fable 5. In collaboration with Chetaslua

GPT-5.6 vs GPT-5.5 on my custom spaceship prompt. I gave both models the exact same custom prompt. This is also the same prompt I previously gave to Fable 5. For context, GPT-5.6 Pro worked for 87 minutes, while GPT-5.5 Extra High worked for 34 minutes and 42 seconds. As I’ve said before, based on great authority GPT-5.6 will be an incremental/soldi improvement over GPT-5.5, not a “Fable killer.” My rough expectation has been that it would trade blows with Fable 5 on some benchmarks, maybe win around half depending on the category, but not clearly surpass it overall. And again fable five will have bigger model smell, but this was expected. After testing this coding output, that view feels pretty accurate. GPT-5.6 is clearly better than GPT-5.5 in several visual areas. The lighting, shading, chairs, object details, and exterior of the spaceship looked noticeably stronger. The scene was also easier to test. I do want to give GPT-5.5 credit though. It built out the rooms much much better and the planets looked better than GPT-5.6’s. It was also interesting that both GPT-5.5 and GPT-5.6 produced better-looking planets than Fable 5 in this specific test. The downside with GPT-5.5 was stability. The game was much glitchier and harder to test compared to GPT-5.6. But when it comes to the core of the demo, which is the spaceship itself, Fable 5 still beat both models pretty comfortably. GPT-5.6 is impressive, but from this test, it looks exactly like what I expected which was a meaningful incremental improvement over GPT-5.5, at least for indie game demos, but not something that replaces Fable 5. In collaboration with Chetaslua

Chris

249,508 次观看 • 28 天前

meta muse spark 1.1 vs gpt 5.6 sol vs fable 5 vs grok 4.5 meta recently dropped muse spark 1.1 – a multimodal reasoning model from meta superintelligence labs built for agentic tasks. key facts: • 1m token context with active self-management – the model compacts its own history and keeps only the steps needed for later work • trained to orchestrate multi-agent systems: as main agent it plans and delegates to parallel subagents, as subagent it sticks to its job and knows when to escalate back • computer use trained to pick between scripting and clicking – writes automation when it's faster, clicks when it's simpler, batches actions per step • first public api from meta: the meta model api is now in preview • benchmarks: sweeps the agent column – mcp atlas 88.1 (opus 4.8: 82.2), jobbench 54.7 (opus: 48.4), humanity's last exam 62.1 (1st). loses coding – deepswe 1.1 53.3 vs gpt 5.5's 67.0, swe bench pro 61.5 vs opus's 69.2 our test – 3 prompts, single-file html, three.js, fully procedural, no assets: 1. norwegian house cantilevered over a fjord in a snowstorm – transmissive glass wall, fully modelled interior 2. beijing siheyuan courtyard house in dawn fog – instanced roof tiles, dougong brackets, glowing paper windows 3. new mexico adobe pueblo in an approaching dust storm – deep window reveals, windward grit accumulation we ran the test on AI/ML API platform results: - cost #1 muse spark 1.1 – $0.20 #2 grok 4.5 – $0.51 #3 gpt 5.6 sol – $1.93 #4 fable 5 – ~$5.20 - output tokens #1 muse spark 1.1 – 41,868 #2 gpt 5.6 sol – 49,139 #3 grok 4.5 – 64,954 #4 fable 5 – 81,849 - lines of code #1 muse spark 1.1 – 1,799 #2 gpt 5.6 sol – 2,377 #3 fable 5 – 3,088 #4 grok 4.5 – 4,216 observations: • muse spark is the cheapest of the four by a wide margin – 2.5x under grok, ~26x under fable per run. output quality tracks the price • only 7.4% of its output tokens are reasoning (3,104 of 41,868) – the model barely thinks before writing. economic, not pedantic: it commits to the first plan and ships it • the low loc is not compression, it's omission – all three prompts demanded instancing, muse spark delivered it in one muse spark's code quality – reviewed by fable 5: upsides: 1. all three files run 2. the adobe grit effect is legit – shader injection via onbeforecompile, windward faces detect storm direction through a normal-dot-wind term and darken procedurally 3. the fjord glass is real meshphysicalmaterial with transmission and ior, not a transparent quad 4. the siheyuan properly instances barrel tiles, dougong blocks and courtyard pavers downsides: 1. in the fjord file the strafe vector is negated – press a, you move right; press d, you move left. exactly the key mix-up we kept hitting with this model 2. all three files ship the model's self-doubt as comments: "// actually yaw orientation: need correct" sits above a direction vector that gets computed, abandoned and recomputed – dead vectors allocated every frame, 60 times a second 3. the siheyuan registers two separate keydown listeners, one containing an empty if-block 4. snow "accumulation" on the norway roof is a sine wobble on a scale value, not accumulation 5. "instanced snow" became 3,500 plain points. zero dispose calls anywhere pattern: minimal reasoning, minimal code, minimal price. it nails the flashy requirements – shaders, transmissive glass – and quietly drops the boring ones: instancing, controls, cleanup. you get a demo that mostly runs and a control scheme you can't trust follow thehype. for 24/7 ai news, analysis and breakdowns

meta muse spark 1.1 vs gpt 5.6 sol vs fable 5 vs grok 4.5 meta recently dropped muse spark 1.1 – a multimodal reasoning model from meta superintelligence labs built for agentic tasks. key facts: • 1m token context with active self-management – the model compacts its own history and keeps only the steps needed for later work • trained to orchestrate multi-agent systems: as main agent it plans and delegates to parallel subagents, as subagent it sticks to its job and knows when to escalate back • computer use trained to pick between scripting and clicking – writes automation when it's faster, clicks when it's simpler, batches actions per step • first public api from meta: the meta model api is now in preview • benchmarks: sweeps the agent column – mcp atlas 88.1 (opus 4.8: 82.2), jobbench 54.7 (opus: 48.4), humanity's last exam 62.1 (1st). loses coding – deepswe 1.1 53.3 vs gpt 5.5's 67.0, swe bench pro 61.5 vs opus's 69.2 our test – 3 prompts, single-file html, three.js, fully procedural, no assets: 1. norwegian house cantilevered over a fjord in a snowstorm – transmissive glass wall, fully modelled interior 2. beijing siheyuan courtyard house in dawn fog – instanced roof tiles, dougong brackets, glowing paper windows 3. new mexico adobe pueblo in an approaching dust storm – deep window reveals, windward grit accumulation we ran the test on AI/ML API platform results: - cost #1 muse spark 1.1 – $0.20 #2 grok 4.5 – $0.51 #3 gpt 5.6 sol – $1.93 #4 fable 5 – ~$5.20 - output tokens #1 muse spark 1.1 – 41,868 #2 gpt 5.6 sol – 49,139 #3 grok 4.5 – 64,954 #4 fable 5 – 81,849 - lines of code #1 muse spark 1.1 – 1,799 #2 gpt 5.6 sol – 2,377 #3 fable 5 – 3,088 #4 grok 4.5 – 4,216 observations: • muse spark is the cheapest of the four by a wide margin – 2.5x under grok, ~26x under fable per run. output quality tracks the price • only 7.4% of its output tokens are reasoning (3,104 of 41,868) – the model barely thinks before writing. economic, not pedantic: it commits to the first plan and ships it • the low loc is not compression, it's omission – all three prompts demanded instancing, muse spark delivered it in one muse spark's code quality – reviewed by fable 5: upsides: 1. all three files run 2. the adobe grit effect is legit – shader injection via onbeforecompile, windward faces detect storm direction through a normal-dot-wind term and darken procedurally 3. the fjord glass is real meshphysicalmaterial with transmission and ior, not a transparent quad 4. the siheyuan properly instances barrel tiles, dougong blocks and courtyard pavers downsides: 1. in the fjord file the strafe vector is negated – press a, you move right; press d, you move left. exactly the key mix-up we kept hitting with this model 2. all three files ship the model's self-doubt as comments: "// actually yaw orientation: need correct" sits above a direction vector that gets computed, abandoned and recomputed – dead vectors allocated every frame, 60 times a second 3. the siheyuan registers two separate keydown listeners, one containing an empty if-block 4. snow "accumulation" on the norway roof is a sine wobble on a scale value, not accumulation 5. "instanced snow" became 3,500 plain points. zero dispose calls anywhere pattern: minimal reasoning, minimal code, minimal price. it nails the flashy requirements – shaders, transmissive glass – and quietly drops the boring ones: instancing, controls, cleanup. you get a demo that mostly runs and a control scheme you can't trust follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

132,390 次观看 • 6 天前

I tested GLM 5.2 on three real builds. A horror house game. A 3D stealth game. A Remotion marketing video. The results are exactly what I expect from Chinese open source coding models right now. The horror house has broken game logic. Collect all three keys, complete the objective, and the front door still won't open. The stealth game shipped with lighting so dark you can't see your own character. The Remotion video is rough. GLM 5.2 is faster than 5.1 and 10x cheaper than the frontier. But Opus 4.8 and GPT 5.5 are still in a different league for serious coding work. Chinese models keep missing the attention to detail that matters. I MISS FABLE 5 ALREADY. Full review live now.

I tested GLM 5.2 on three real builds. A horror house game. A 3D stealth game. A Remotion marketing video. The results are exactly what I expect from Chinese open source coding models right now. The horror house has broken game logic. Collect all three keys, complete the objective, and the front door still won't open. The stealth game shipped with lighting so dark you can't see your own character. The Remotion video is rough. GLM 5.2 is faster than 5.1 and 10x cheaper than the frontier. But Opus 4.8 and GPT 5.5 are still in a different league for serious coding work. Chinese models keep missing the attention to detail that matters. I MISS FABLE 5 ALREADY. Full review live now.

BridgeMind

24,527 次观看 • 1 个月前

Before Claude Fable 5 got banned, I turned all my fine-tuning research and experiments into a product: (A CLI for data generation to fine-tune AI models). Finetuner uses Codex 5.5 or Opus 4.8 as the orchestrator and Chinese models (DeepSeek v4 Pro, Kimi K2.7, MiMo v2.5, etc.) to generate dataset rows. This system is unique because these are not Python-script-paraphrased datasets. Each and every row is handcrafted by Chinese models based on the orchestrator model's instructions. Now anyone can generate datasets to fine-tune small language models (1B to 30B models). I achieved 10x lower costs and 5x better dataset quality. Releasing the product in a few days. (I'll open-source the skills.)

Before Claude Fable 5 got banned, I turned all my fine-tuning research and experiments into a product: (A CLI for data generation to fine-tune AI models). Finetuner uses Codex 5.5 or Opus 4.8 as the orchestrator and Chinese models (DeepSeek v4 Pro, Kimi K2.7, MiMo v2.5, etc.) to generate dataset rows. This system is unique because these are not Python-script-paraphrased datasets. Each and every row is handcrafted by Chinese models based on the orchestrator model's instructions. Now anyone can generate datasets to fine-tune small language models (1B to 30B models). I achieved 10x lower costs and 5x better dataset quality. Releasing the product in a few days. (I'll open-source the skills.)

CJ Zafir

39,716 次观看 • 1 个月前

I sat down with Howie Liu, the CEO of Airtable ($500M+ revenue, 1 billion in the bank) and asked him: is there really 1 trillion up for grabs in AI agents? His answer: it's way more than that. It's the entire GDP of white collar labor. Tens of trillions. Here's what stood out: 1. Howie runs 30 Claude Code instances in parallel on HyperAgent. Each one is coupled to a browser, fully autonomous. They review each other's PRs. That's how the CEO of a $10 billion company develops software right now. 2. He wrote his most recent board memo with AI agents. His best investors told him it was the best memo he'd ever written. It cost him $150 in tokens and 10x less time. 3. His take on why people aren't building: they're still using agents like chatbots. They ask "who's going to win the next election" instead of giving it a real multi-hour task. Using is believing. You have to spend a full weekend going deep. 4. AI agents are at less than 10% penetration in most industries. Software engineering is at 50% but even that's an overestimate because most devs are still in "tab autocomplete" mode. The frontier has moved way past that. 5. He revealed HyperAgent. Think of it as the visual agent builder that gives you a low floor and a high ceiling. You can prototype fast and also scale to running serious operations with a fleet of agents. 6. Howie's philosophy/POV: HyperAgent is to agents what the iPhone was to computing. The power was already there. The accessibility is what changes everything. Good news Howie is giving $1,000 in free HyperAgent credits to the first 1,000 people who sign up. $1 million committed to listeners The Startup Ideas Podcast (SIP) 🧃. You get Opus, frontier models, real agent workflows. You just gotta click the link in the description of the YT vid (share this with a friend to give them the $1000 too before it runs out!) episode is live on The Startup Ideas Podcast (SIP) 🧃 and thanks to Howie for supporting the community/channel. Howie Liu is rooting for you to build a $100 million company with less than 5 employees. So am I. watch

I sat down with Howie Liu, the CEO of Airtable ($500M+ revenue, 1 billion in the bank) and asked him: is there really 1 trillion up for grabs in AI agents? His answer: it's way more than that. It's the entire GDP of white collar labor. Tens of trillions. Here's what stood out: 1. Howie runs 30 Claude Code instances in parallel on HyperAgent. Each one is coupled to a browser, fully autonomous. They review each other's PRs. That's how the CEO of a $10 billion company develops software right now. 2. He wrote his most recent board memo with AI agents. His best investors told him it was the best memo he'd ever written. It cost him $150 in tokens and 10x less time. 3. His take on why people aren't building: they're still using agents like chatbots. They ask "who's going to win the next election" instead of giving it a real multi-hour task. Using is believing. You have to spend a full weekend going deep. 4. AI agents are at less than 10% penetration in most industries. Software engineering is at 50% but even that's an overestimate because most devs are still in "tab autocomplete" mode. The frontier has moved way past that. 5. He revealed HyperAgent. Think of it as the visual agent builder that gives you a low floor and a high ceiling. You can prototype fast and also scale to running serious operations with a fleet of agents. 6. Howie's philosophy/POV: HyperAgent is to agents what the iPhone was to computing. The power was already there. The accessibility is what changes everything. Good news Howie is giving $1,000 in free HyperAgent credits to the first 1,000 people who sign up. $1 million committed to listeners The Startup Ideas Podcast (SIP) 🧃. You get Opus, frontier models, real agent workflows. You just gotta click the link in the description of the YT vid (share this with a friend to give them the $1000 too before it runs out!) episode is live on The Startup Ideas Podcast (SIP) 🧃 and thanks to Howie for supporting the community/channel. Howie Liu is rooting for you to build a $100 million company with less than 5 employees. So am I. watch

GREG ISENBERG

39,210,670 次观看 • 2 个月前

One of the smartest things you can do with Fable 5 right now: Re-create your AI second brain to log all your business ideas, personal context, and important data. The first time I built an AI second brain was with Opus, but I recently re-created it with Fable 5, and it blew my mind. Here's exactly how to get started: Step 1. Set up your Obsidian vault Download Obsidian from Obsidian dot md if you haven't already. Then, go ahead and create a clean vault with your most important folders. For example: /ideas → business ideas, content angles, random thoughts /context → who you are, your business, your goals, your stack /data → important numbers, portfolios, metrics /log → daily entries, decisions, lessons learned This is your database. Everything Fable reads lives here. Step 2. Connect Fable 5 to your vault I like this Claude Code prompt: "/goal connect to my Obsidian vault at [path] and act as my second brain orchestrator. Read everything in /context before every session. Log anything new I tell you to /log with today's date." Fable now reads your vault before it answers anything - it knows your business, your goals, your history. Step 3. Build the self-update habit Every time you have an idea, a decision, or a lesson, tell Fable: "Log this to my second brain: [thought]" Step 4. Start querying it You can start sending prompts like: → "What are the most common themes across my last 30 ideas?" → "Based on my context, what should I be prioritising this week?" → "What decisions have I made about my content strategy so far?" Opus was good at this, but Fable is on another level. I feel the depth of reasoning it brings to your data is genuinely unlike anything I've used before. Some might argue it's a bit of overkill to use Fable for a simple second-brain setup, but if you have the means, it's 100% worth it. Build this once, and it'll compound forever.

One of the smartest things you can do with Fable 5 right now: Re-create your AI second brain to log all your business ideas, personal context, and important data. The first time I built an AI second brain was with Opus, but I recently re-created it with Fable 5, and it blew my mind. Here's exactly how to get started: Step 1. Set up your Obsidian vault Download Obsidian from Obsidian dot md if you haven't already. Then, go ahead and create a clean vault with your most important folders. For example: /ideas → business ideas, content angles, random thoughts /context → who you are, your business, your goals, your stack /data → important numbers, portfolios, metrics /log → daily entries, decisions, lessons learned This is your database. Everything Fable reads lives here. Step 2. Connect Fable 5 to your vault I like this Claude Code prompt: "/goal connect to my Obsidian vault at [path] and act as my second brain orchestrator. Read everything in /context before every session. Log anything new I tell you to /log with today's date." Fable now reads your vault before it answers anything - it knows your business, your goals, your history. Step 3. Build the self-update habit Every time you have an idea, a decision, or a lesson, tell Fable: "Log this to my second brain: [thought]" Step 4. Start querying it You can start sending prompts like: → "What are the most common themes across my last 30 ideas?" → "Based on my context, what should I be prioritising this week?" → "What decisions have I made about my content strategy so far?" Opus was good at this, but Fable is on another level. I feel the depth of reasoning it brings to your data is genuinely unlike anything I've used before. Some might argue it's a bit of overkill to use Fable for a simple second-brain setup, but if you have the means, it's 100% worth it. Build this once, and it'll compound forever.

Miles Deutscher

82,831 次观看 • 3 天前

HERMES AGENT NOW RUNS A FULL BUSINESS FOR $8/MONTH. CONTENT. CODE. INBOX. ALL AGENTS. EVERYTHING WHILE YOU SLEEP. one user built this in one session. zero terminal commands. browser only. the team: → content agent: finds video topics, writes first drafts, helps with animations → code builder: writes website and SaaS code → ops agent: organizes and answers email total setup cost: $1. monthly runtime: $8 VPS + cheap model API. the trick is matching model to task. frontier models (Opus 4.8, GPT-5.5) for complex /goals only. $5/million tokens. mid-tier (Claude Sonnet, Haiku, Kimiko 2.6) for daily content and research work. budget (DeepSeek V4 Flash, MiniMax 2.7, Gemma 4, Qwen 3) for cron jobs and routine. run the initial setup on Opus. switch to DeepSeek for everything after. your bill drops 90% with zero quality loss on routine work. the agent team picks up tasks from kanban. each completed task creates a skill. week 3 the same workflows run faster than week 1. full breakdown on Hermes Os in the article 👇

HERMES AGENT NOW RUNS A FULL BUSINESS FOR $8/MONTH. CONTENT. CODE. INBOX. ALL AGENTS. EVERYTHING WHILE YOU SLEEP. one user built this in one session. zero terminal commands. browser only. the team: → content agent: finds video topics, writes first drafts, helps with animations → code builder: writes website and SaaS code → ops agent: organizes and answers email total setup cost: $1. monthly runtime: $8 VPS + cheap model API. the trick is matching model to task. frontier models (Opus 4.8, GPT-5.5) for complex /goals only. $5/million tokens. mid-tier (Claude Sonnet, Haiku, Kimiko 2.6) for daily content and research work. budget (DeepSeek V4 Flash, MiniMax 2.7, Gemma 4, Qwen 3) for cron jobs and routine. run the initial setup on Opus. switch to DeepSeek for everything after. your bill drops 90% with zero quality loss on routine work. the agent team picks up tasks from kanban. each completed task creates a skill. week 3 the same workflows run faster than week 1. full breakdown on Hermes Os in the article 👇

YanXbt

30,802 次观看 • 1 个月前

BREAKING: GPT-5.6 Sol is out—AND Codex has been merged into ChatGPT Desktop as ChatGPT Codex. This combo model and desktop app harness are the gold-standard for knowledge work in AI. 5.6 is powerful, fast, half the price of Fable, and my default for almost everything. We’ve been testing it internally Every 📧 for about a month across coding, writing, design, and knowledge work. Here’s our day-zero vibe check: - An A-tier coder—but it’s not Fable. Sol scored 56/100 on our Senior Engineer benchmark compared to a 91 for Fable. I think the 56/100 undersells it, it's an excellent implementor, and very smart. But Fable just writes conceptually cleaner code and works better at the top end of task complexity. PRO-TIP: Use GPT-5.6 as Fable's subagent for the most goated combo in AI coding. - The best writer of the frontier models. It’s clearer and more concise than Fable or Opus 4.8, without the overexplaining or weird private language. It can one-shot marketing emails, help you workshop taglines, and explain complex concepts clearly. It's also super fast, which makes it easy to collaborate with. - Design is better, but not top-tier. It has noticeably more taste than 5.5, but Fable and Opus 4.8 are still playing at a different level. See examples in the video and vibe check below. - The real leap is knowledge work. Sol is the first model I’ve trusted to run whole loops of knowledge work—not just help with individual tasks. I use it to process email, surface decisions from meetings and Slack, find job candidates, scan Facebook Marketplace for furniture, and log my meals. It has shifted my job from doing the work to tending the system that does it. - The merged app is fine. I was extremely worried about this because I love the Codex app. OpenAI was caught in an interesting position: How to make an agent orchestration app for regular ChatGPT consumers, coders, and businesses all in one app. They now split the interface between ChatGPT Work and ChatGPT Codex. They're basically the same except Work hides code. And "Chat" has been demoted to 2nd tier status for quick questions in either one. It's not a big leap, but it's not a huge setback either. And it remains my favorite of the desktop agent orchestration apps. Verdict: If I really had to put my finger on it, I'd say Fable has way more big model smell. But that means it's a skill in itself to get value out of it—99% of people are still not there yet. GPT-5.6 is almost as powerful, but is easy to use, fast, and relatively cheap. It should give you an early sense of where model work is going. Full Every 📧 Vibe Check:

BREAKING: GPT-5.6 Sol is out—AND Codex has been merged into ChatGPT Desktop as ChatGPT Codex. This combo model and desktop app harness are the gold-standard for knowledge work in AI. 5.6 is powerful, fast, half the price of Fable, and my default for almost everything. We’ve been testing it internally Every 📧 for about a month across coding, writing, design, and knowledge work. Here’s our day-zero vibe check: - An A-tier coder—but it’s not Fable. Sol scored 56/100 on our Senior Engineer benchmark compared to a 91 for Fable. I think the 56/100 undersells it, it's an excellent implementor, and very smart. But Fable just writes conceptually cleaner code and works better at the top end of task complexity. PRO-TIP: Use GPT-5.6 as Fable's subagent for the most goated combo in AI coding. - The best writer of the frontier models. It’s clearer and more concise than Fable or Opus 4.8, without the overexplaining or weird private language. It can one-shot marketing emails, help you workshop taglines, and explain complex concepts clearly. It's also super fast, which makes it easy to collaborate with. - Design is better, but not top-tier. It has noticeably more taste than 5.5, but Fable and Opus 4.8 are still playing at a different level. See examples in the video and vibe check below. - The real leap is knowledge work. Sol is the first model I’ve trusted to run whole loops of knowledge work—not just help with individual tasks. I use it to process email, surface decisions from meetings and Slack, find job candidates, scan Facebook Marketplace for furniture, and log my meals. It has shifted my job from doing the work to tending the system that does it. - The merged app is fine. I was extremely worried about this because I love the Codex app. OpenAI was caught in an interesting position: How to make an agent orchestration app for regular ChatGPT consumers, coders, and businesses all in one app. They now split the interface between ChatGPT Work and ChatGPT Codex. They're basically the same except Work hides code. And "Chat" has been demoted to 2nd tier status for quick questions in either one. It's not a big leap, but it's not a huge setback either. And it remains my favorite of the desktop agent orchestration apps. Verdict: If I really had to put my finger on it, I'd say Fable has way more big model smell. But that means it's a skill in itself to get value out of it—99% of people are still not there yet. GPT-5.6 is almost as powerful, but is easy to use, fast, and relatively cheap. It should give you an early sense of where model work is going. Full Every 📧 Vibe Check:

Dan Shipper 📧

144,276 次观看 • 8 天前

What can you do now to protect yourself for the winter 2025-26 COVlD wave? 5 tips: 1) Take care of business now, e.g. dental apps 2) Boost up (no script required in the US) 3) Wear a well-fitting high-quality mask (N95/elastomeric, no gaps at the cheeks, nose, or chin) 4) Test before gatherings to reduce the risk of forward transmission (and daily if sick, assume positive in the meantime). FYI, that "positive test" is 3D printed, never had a positive due to following these tips. 5) Clean the air (outdoors, windows open, HVAC on, HEPA, DIY air cleaners, and go bigger than that tiny HEPA shown) We will post updated surveillance data shortly on the PMC website. Also, I'll be on FOX8 tonight from 5-6pm Central.

What can you do now to protect yourself for the winter 2025-26 COVlD wave? 5 tips: 1) Take care of business now, e.g. dental apps 2) Boost up (no script required in the US) 3) Wear a well-fitting high-quality mask (N95/elastomeric, no gaps at the cheeks, nose, or chin) 4) Test before gatherings to reduce the risk of forward transmission (and daily if sick, assume positive in the meantime). FYI, that "positive test" is 3D printed, never had a positive due to following these tips. 5) Clean the air (outdoors, windows open, HVAC on, HEPA, DIY air cleaners, and go bigger than that tiny HEPA shown) We will post updated surveillance data shortly on the PMC website. Also, I'll be on FOX8 tonight from 5-6pm Central.

Mike Hoerger, PhD MSCR MBA

31,649 次观看 • 8 个月前

$On August 11, the supply of SpaceX shares jumps from 5% to 25% in a single day. That's 5 times more shares hitting the market overnight, from insiders who bought in years ago at a fraction of today's price. This is what happened to Facebook in 2012. After their lockup expired, the stock fell about 50% in just a few weeks. The business was growing fast at the time. The drop came purely from supply hitting the market faster than buyers could absorb. Most people holding SpaceX have no plan for August 11. Three steps to prepare for it now: 1. Print the unlock schedule and know every date new shares hit the market. 2. Check if SpaceX is more than 10% of your portfolio. 3. Decide your exit price now, before emotion kicks in. I'm watching closely. Definitely not buying yet.$

On August 11, the supply of SpaceX shares jumps from 5% to 25% in a single day. That's 5 times more shares hitting the market overnight, from insiders who bought in years ago at a fraction of today's price. This is what happened to Facebook in 2012. After their lockup expired, the stock fell about 50% in just a few weeks. The business was growing fast at the time. The drop came purely from supply hitting the market faster than buyers could absorb. Most people holding SpaceX have no plan for August 11. Three steps to prepare for it now: 1. Print the unlock schedule and know every date new shares hit the market. 2. Check if SpaceX is more than 10% of your portfolio. 3. Decide your exit price now, before emotion kicks in. I'm watching closely. Definitely not buying yet.

Felix Prehn 🐶

167,451 次观看 • 22 天前

LaunchDay #4 starts now 🥳 I met with 11 indie makers from all over the world, each with a unique story: - Arthur is just 20 years old, but already runs a 5-people agency. He recently started building his own products on the side - Daniel left Malta for the USA a few years ago to work for AWS, then became one of the most successful indie makers in the world - Lera quit her regular design job to build her own productized service, and now has a team of 5 people working for her - ...and 8 more amazing people Today they're all on giving a huge discount on their products, for 48 hours only ⚡ So if you love indie makers, please go check them out, and retweet this post for reach. Time to come together and help each other out 💙

LaunchDay #4 starts now 🥳 I met with 11 indie makers from all over the world, each with a unique story: - Arthur is just 20 years old, but already runs a 5-people agency. He recently started building his own products on the side - Daniel left Malta for the USA a few years ago to work for AWS, then became one of the most successful indie makers in the world - Lera quit her regular design job to build her own productized service, and now has a team of 5 people working for her - ...and 8 more amazing people Today they're all on giving a huge discount on their products, for 48 hours only ⚡ So if you love indie makers, please go check them out, and retweet this post for reach. Time to come together and help each other out 💙

Dagobert

20,464 次观看 • 11 个月前