Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

We worked with Ramp Labs to train Fast Ask using Lab A small RL-trained subagent that helps the Ramp Sheets agent find answers in spreadsheets. The resulting FastAsk model outperformed Opus 4.6, while obtaining Haiku-level speeds at even lower costs.

Prime Intellect

65,738 subscribers

55,114 görüntüleme • 3 ay önce •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

Why did a spend management platform start its own AI research lab? Karim Atiyeh describes Ramp Labs as a “collection of crazy ambitious people that don’t like to be told that something is impossible”. It started about a year ago and has since worked on projects like building its own production-grounded coding benchmark “Ramp SWE-Bench”, putting Claude Code in RollerCoaster Tycoon, and a mechanistic interpretability playground. The Redpoint team spent the day with Ramp Labs to answer questions like: ▪️ How did Ramp Labs get started? ▪️ How do they decide what research to pursue? ▪️ The long-term vision for Ramp Labs

Why did a spend management platform start its own AI research lab? Karim Atiyeh describes Ramp Labs as a “collection of crazy ambitious people that don’t like to be told that something is impossible”. It started about a year ago and has since worked on projects like building its own production-grounded coding benchmark “Ramp SWE-Bench”, putting Claude Code in RollerCoaster Tycoon, and a mechanistic interpretability playground. The Redpoint team spent the day with Ramp Labs to answer questions like: ▪️ How did Ramp Labs get started? ▪️ How do they decide what research to pursue? ▪️ The long-term vision for Ramp Labs

Redpoint

115,895 görüntüleme • 13 gün önce

Introducing Ramp Tour Guide: an AI Agent that can show you how to do anything on Ramp! Today, we'd like to share a sneak peek of Ramp's near future. As Ramp grows in functionality, we want to make all of it easily accessible to all of our customers. To do that, we're demoing a first-of-its-kind AI agent that can show and tell people how to accomplish anything with Ramp. The Ramp Tour Guide knows Ramp inside and out. You can ask it how to do something on our platform and it'll walk you through every step of the way.

Introducing Ramp Tour Guide: an AI Agent that can show you how to do anything on Ramp! Today, we'd like to share a sneak peek of Ramp's near future. As Ramp grows in functionality, we want to make all of it easily accessible to all of our customers. To do that, we're demoing a first-of-its-kind AI agent that can show and tell people how to accomplish anything with Ramp. The Ramp Tour Guide knows Ramp inside and out. You can ask it how to do something on our platform and it'll walk you through every step of the way.

Ramp

473,792 görüntüleme • 2 yıl önce

we created a claude agent that helps you build a fintech app from scratch, with just one simple prompt in claude code the subagent file was refined by opus 4.5, which distilled it down to a really simple, easy-to-follow guide even for humans the future is here, go build!

we created a claude agent that helps you build a fintech app from scratch, with just one simple prompt in claude code the subagent file was refined by opus 4.5, which distilled it down to a really simple, easy-to-follow guide even for humans the future is here, go build!

joyce

452,747 görüntüleme • 7 ay önce

Jensen is showing off the tender hotpot meat in Asia. Hopefully, due to deep partnerships with the Asian supply chain, the VR200 NVL144 new product introduction ramp will go off with much fewer major bumps than the GB200 NVL72 ramp. The main potential risks to the VR200 NVL144 ramp are: 1. VR200 has upgraded its TGP from 1800W to ~2200-2300W in order to widen the FLOP gap against MI450X. 2. VR200 has upgraded its memory bandwidth from 13TB/s to 20TB/s in order to match MI450X. VR200 does this by using higher-bin HBM 3. VR200 is potentially using 448G BiDirectional SerDes, where it can achieve 224G RX & 224G TX simultaneously on the same copper cable in parallel at the same time. Versus on GB200 NVL72 backplane, each direction requires a dedicated copper cable All of these push the limits of physics and could potentially pose a risk to the ramp, but we are hopeful that Jensen will work with the Asian supply chain to ensure a smoother ramp since this is his 3rd time doing a rack-scale computing ramp.

Jensen is showing off the tender hotpot meat in Asia. Hopefully, due to deep partnerships with the Asian supply chain, the VR200 NVL144 new product introduction ramp will go off with much fewer major bumps than the GB200 NVL72 ramp. The main potential risks to the VR200 NVL144 ramp are: 1. VR200 has upgraded its TGP from 1800W to ~2200-2300W in order to widen the FLOP gap against MI450X. 2. VR200 has upgraded its memory bandwidth from 13TB/s to 20TB/s in order to match MI450X. VR200 does this by using higher-bin HBM 3. VR200 is potentially using 448G BiDirectional SerDes, where it can achieve 224G RX & 224G TX simultaneously on the same copper cable in parallel at the same time. Versus on GB200 NVL72 backplane, each direction requires a dedicated copper cable All of these push the limits of physics and could potentially pose a risk to the ramp, but we are hopeful that Jensen will work with the Asian supply chain to ensure a smoother ramp since this is his 3rd time doing a rack-scale computing ramp.

SemiAnalysis

123,347 görüntüleme • 9 ay önce

GLM 4.6 runs quite fast on an M3 Ultra with mlx-lm even at higher precision. Pretty remarkable that it benchmarks competitive to the just-released Sonnet 4.5. Hope those benchmarks hold-up in day-to-day use. Here's a run using 5.5 bpw quantized model, generating 5.3k tokens at 17+ tok/sec using 244 GB. What prompts should I test?

GLM 4.6 runs quite fast on an M3 Ultra with mlx-lm even at higher precision. Pretty remarkable that it benchmarks competitive to the just-released Sonnet 4.5. Hope those benchmarks hold-up in day-to-day use. Here's a run using 5.5 bpw quantized model, generating 5.3k tokens at 17+ tok/sec using 244 GB. What prompts should I test?

Awni Hannun

68,539 görüntüleme • 10 ay önce

🚨 do you understand what just happened to Microsoft.. Microsoft just dropped seven of its own MAI models, trained from scratch with zero distillation, and said its custom-tuned models already match GPT-tier quality at 10x lower cost. The partner that pays OpenAI's bills is now quietly building the thing that replaces it. - MAI-Thinking-1 hits human-preference parity with Sonnet 4.6 and 53% on SWE Bench Pro, right next to Opus 4.6 - MAI-Code-1-Flash delivers Haiku-class coding at just 5B params and is already shipping inside GitHub Copilot - A custom MAI model tuned for Excel matched GPT-5.4 while running up to 10x more efficiently - Their real weapon isn't the weights, it's that they're already inside everyone's Office, Teams, and VS Code The hard part was never training a good model. The hard part is being the room everyone already works in. Microsoft owns the room.

🚨 do you understand what just happened to Microsoft.. Microsoft just dropped seven of its own MAI models, trained from scratch with zero distillation, and said its custom-tuned models already match GPT-tier quality at 10x lower cost. The partner that pays OpenAI's bills is now quietly building the thing that replaces it. - MAI-Thinking-1 hits human-preference parity with Sonnet 4.6 and 53% on SWE Bench Pro, right next to Opus 4.6 - MAI-Code-1-Flash delivers Haiku-class coding at just 5B params and is already shipping inside GitHub Copilot - A custom MAI model tuned for Excel matched GPT-5.4 while running up to 10x more efficiently - Their real weapon isn't the weights, it's that they're already inside everyone's Office, Teams, and VS Code The hard part was never training a good model. The hard part is being the room everyone already works in. Microsoft owns the room.

BuBBliK

171,865 görüntüleme • 2 ay önce

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 görüntüleme • 3 ay önce

Wow. This is crazy. A developer trained an AI agent in simulation and deployed it onto a real robotic air hockey table using reinforcement learning. This robot can track the puck with millimeter-level accuracy and react in roughly 20 milliseconds, fast enough to challenge even skilled human players. We’re moving from robots that follow programmed rules to machines that learn strategies in simulation and execute them in the physical world.

Wow. This is crazy. A developer trained an AI agent in simulation and deployed it onto a real robotic air hockey table using reinforcement learning. This robot can track the puck with millimeter-level accuracy and react in roughly 20 milliseconds, fast enough to challenge even skilled human players. We’re moving from robots that follow programmed rules to machines that learn strategies in simulation and execute them in the physical world.

SciTech Era

1,585,107 görüntüleme • 2 ay önce

Today we Point Nine are excited to launch the P9 AI Fluency Index. One of the big challenges for many founders right now is to get their organizations fully "AI-pilled". We wanted to create a tool that that helps founders and CEOs benchmark their company's AI fluency and get recommendations on how to get to the next level. The index is grounded in published rubrics and posts by Zapier, Fin, tobi lutke Shopify, Ramp Labs, Anthropic, Jobber, and Bassim Eledath. Huge thanks to them for publishing their valuable insights! Head over to and come back with your report card! :)

Today we Point Nine are excited to launch the P9 AI Fluency Index. One of the big challenges for many founders right now is to get their organizations fully "AI-pilled". We wanted to create a tool that that helps founders and CEOs benchmark their company's AI fluency and get recommendations on how to get to the next level. The index is grounded in published rubrics and posts by Zapier, Fin, tobi lutke Shopify, Ramp Labs, Anthropic, Jobber, and Bassim Eledath. Huge thanks to them for publishing their valuable insights! Head over to and come back with your report card! :)

Christoph Janz 🕊

18,528 görüntüleme • 3 ay önce

J-Cal: Democrats Will Sweep the Midterms if Trump Doesn’t Find an Off Ramp FAST @jason: I do think Trump would only (start a war with Iran) if he had a very high probability of success and an off ramp. However, it's not looking good with the off ramp right now. If the neocons get their way and the people on Polymarket are correct, the sharps who say 57% chance we'll have boots on the ground, I think this is the end of Trump's second term. And if you were to put together the series of mistakes that he's made, and the administration's made, they're really at the heart of why people voted for him. 1) Doing the war that everybody said he should not do. 2) The Epstein Files. 3) The ICE cruelty. 4) Not working for the American working man who doesn't own equities. If this continues for another six months, it's basically going to result in the Democrats doing a clean sweep in the midterms. This is the chart that should be absolutely terrifying. The chances of the Democrats sweeping now is up to 45%. If he doesn't find an off ramp quickly, they're going to lose both houses in the midterms. That's the thing Trump needs to really consider, and I think he will consider it. I think he's going to find an off ramp.

J-Cal: Democrats Will Sweep the Midterms if Trump Doesn’t Find an Off Ramp FAST @jason: I do think Trump would only (start a war with Iran) if he had a very high probability of success and an off ramp. However, it's not looking good with the off ramp right now. If the neocons get their way and the people on Polymarket are correct, the sharps who say 57% chance we'll have boots on the ground, I think this is the end of Trump's second term. And if you were to put together the series of mistakes that he's made, and the administration's made, they're really at the heart of why people voted for him. 1) Doing the war that everybody said he should not do. 2) The Epstein Files. 3) The ICE cruelty. 4) Not working for the American working man who doesn't own equities. If this continues for another six months, it's basically going to result in the Democrats doing a clean sweep in the midterms. This is the chart that should be absolutely terrifying. The chances of the Democrats sweeping now is up to 45%. If he doesn't find an off ramp quickly, they're going to lose both houses in the midterms. That's the thing Trump needs to really consider, and I think he will consider it. I think he's going to find an off ramp.

The All-In Podcast

55,707 görüntüleme • 5 ay önce

Sorry about so many updates, but this is happening so fast that I want to stay on top of it in real time. The signature that we are seeing on satellite is extremely concerning. If it doesn't fade away in another hour or so, let's give it till 10:00 PM Eastern Time, then we need to start looking at whatever really crazy model runs are out there that initiate this storm somewhere in this small area where this convective blow-up is and then ramp it up really quickly, because there's a decent chance that that model run could be correct.

Sorry about so many updates, but this is happening so fast that I want to stay on top of it in real time. The signature that we are seeing on satellite is extremely concerning. If it doesn't fade away in another hour or so, let's give it till 10:00 PM Eastern Time, then we need to start looking at whatever really crazy model runs are out there that initiate this storm somewhere in this small area where this convective blow-up is and then ramp it up really quickly, because there's a decent chance that that model run could be correct.

Bryan Maxwell SW FL WX

139,332 görüntüleme • 11 ay önce

Not every AI interaction is best served with a large frontier model. There is a long tail of trillion-token use cases, from tagging to search, best served by a sub-10b-parameter model that runs in milliseconds and costs many orders of magnitude less than the frontier. We built Freesolo Flash to make training loops like SFT and RL easy and end-to-end completable through your coding agent.

Not every AI interaction is best served with a large frontier model. There is a long tail of trillion-token use cases, from tagging to search, best served by a sub-10b-parameter model that runs in milliseconds and costs many orders of magnitude less than the frontier. We built Freesolo Flash to make training loops like SFT and RL easy and end-to-end completable through your coding agent.

Freesolo

28,186 görüntüleme • 1 ay önce

#WATCH | Bengaluru, Karnataka: Co-founder, Sarvam AI, Pratyush Kumar says, "... It's been three years now. We started off building on top of the Llama model, which was the best open model at that time. Sarvam was the only company outside, or maybe in Asia, that contributed to the Llama 3 Series. We understood how to add language skills into models, and Llama 3 was reasonable in Indian languages. When the first set of GPUs arrived in India, we pre-trained from scratch a 2-billion model and worked with Infosys to make it useful for IT services. Then we recognised the importance of scaling up reinforcement learning and built a reasoning model called Sarvam M on top of Mistral's small model. Over the last year, we pre-trained from scratch in India on 4,000 Hopper Series GPUs a 100-billion-plus-parameter model..."

#WATCH | Bengaluru, Karnataka: Co-founder, Sarvam AI, Pratyush Kumar says, "... It's been three years now. We started off building on top of the Llama model, which was the best open model at that time. Sarvam was the only company outside, or maybe in Asia, that contributed to the Llama 3 Series. We understood how to add language skills into models, and Llama 3 was reasonable in Indian languages. When the first set of GPUs arrived in India, we pre-trained from scratch a 2-billion model and worked with Infosys to make it useful for IT services. Then we recognised the importance of scaling up reinforcement learning and built a reasoning model called Sarvam M on top of Mistral's small model. Over the last year, we pre-trained from scratch in India on 4,000 Hopper Series GPUs a 100-billion-plus-parameter model..."

ANI

162,971 görüntüleme • 21 gün önce

OpenClaw meets RL! OpenClaw Agents adapt through memory files and skills, but the base model weights never actually change. OpenClaw-RL solves this! It wraps a self-hosted model as an OpenAI-compatible API, intercepts live conversations from OpenClaw, and trains the policy in the background using RL. The architecture is fully async. This means serving, reward scoring, and training all run in parallel. Once done, weights get hot-swapped after every batch while the agent keeps responding. Currently, it has two training modes: - Binary RL (GRPO): A process reward model scores each turn as good, bad, or neutral. That scalar reward drives policy updates via a PPO-style clipped objective. - On-Policy Distillation: When concrete corrections come in like "you should have checked that file first," it uses that feedback as a richer, directional training signal at the token level. When to use OpenClaw-RL? To be fair, a lot of agent behavior can already be improved through better memory and skill design. OpenClaw's existing skill ecosystem and community-built self-improvement skills handle a wide range of use cases without touching model weights at all. If the agent keeps forgetting preferences, that's a memory problem. And if it doesn't know how to handle a specific workflow, that's a skill problem. Both are solvable at the prompt and context layer. Where RL becomes interesting is when the failure pattern lives deeper in the model's reasoning itself. Things like consistently poor tool selection order, weak multi-step planning, or failing to interpret ambiguous instructions the way a specific user intends. Research on agentic RL (like ARTIST and Agent-R1) has shown that these behavioral patterns hit a ceiling with prompt-based approaches alone, especially in complex multi-turn tasks where the model needs to recover from tool failures or adapt its strategy mid-execution. That's the layer OpenClaw-RL targets, and it's a meaningful distinction from what OpenClaw offers. I have shared the repo in the replies!

OpenClaw meets RL! OpenClaw Agents adapt through memory files and skills, but the base model weights never actually change. OpenClaw-RL solves this! It wraps a self-hosted model as an OpenAI-compatible API, intercepts live conversations from OpenClaw, and trains the policy in the background using RL. The architecture is fully async. This means serving, reward scoring, and training all run in parallel. Once done, weights get hot-swapped after every batch while the agent keeps responding. Currently, it has two training modes: - Binary RL (GRPO): A process reward model scores each turn as good, bad, or neutral. That scalar reward drives policy updates via a PPO-style clipped objective. - On-Policy Distillation: When concrete corrections come in like "you should have checked that file first," it uses that feedback as a richer, directional training signal at the token level. When to use OpenClaw-RL? To be fair, a lot of agent behavior can already be improved through better memory and skill design. OpenClaw's existing skill ecosystem and community-built self-improvement skills handle a wide range of use cases without touching model weights at all. If the agent keeps forgetting preferences, that's a memory problem. And if it doesn't know how to handle a specific workflow, that's a skill problem. Both are solvable at the prompt and context layer. Where RL becomes interesting is when the failure pattern lives deeper in the model's reasoning itself. Things like consistently poor tool selection order, weak multi-step planning, or failing to interpret ambiguous instructions the way a specific user intends. Research on agentic RL (like ARTIST and Agent-R1) has shown that these behavioral patterns hit a ceiling with prompt-based approaches alone, especially in complex multi-turn tasks where the model needs to recover from tool failures or adapt its strategy mid-execution. That's the layer OpenClaw-RL targets, and it's a meaningful distinction from what OpenClaw offers. I have shared the repo in the replies!

Avi Chawla

138,769 görüntüleme • 5 ay önce

If you want to understand why RL struggles with long-horizon agent tasks, this is a good explanation. The core issue is that sparse rewards throw away most of the useful information in the trajectory. GEPA tries to learn from the trajectory itself, using reflection in text space, instead of only optimizing on the final reward. GEPA generates textual critiques of trajectories, proposes prompt edits, and then selects updates along a Pareto frontier between exploration and exploitation instead. Instead of collapsing everything into one reward number, it keeps more information about why a run failed and uses that to make legible changes. It’ll be interesting to see where this goes when people combine that kind of trajectory-level reflection with RL, using RL for optimization while preserving a much richer signal about why the agent succeeded or failed.

If you want to understand why RL struggles with long-horizon agent tasks, this is a good explanation. The core issue is that sparse rewards throw away most of the useful information in the trajectory. GEPA tries to learn from the trajectory itself, using reflection in text space, instead of only optimizing on the final reward. GEPA generates textual critiques of trajectories, proposes prompt edits, and then selects updates along a Pareto frontier between exploration and exploitation instead. Instead of collapsing everything into one reward number, it keeps more information about why a run failed and uses that to make legible changes. It’ll be interesting to see where this goes when people combine that kind of trajectory-level reflection with RL, using RL for optimization while preserving a much richer signal about why the agent succeeded or failed.

Ben Cohen

20,006 görüntüleme • 3 ay önce

🇺🇸🇮🇷 A nuclear submarine is sitting off Cyprus loaded with 20 nuclear-armed missiles while Iran and Israel trade fire. Retired Colonel Anthony Aguilar just walked through exactly how fast this spirals out of anyone's control. "We are now in another level of escalatory territory that is out of Donald Trump's control." The off-ramp isn't just narrow. It may not exist. @AnthonyAgu88102

🇺🇸🇮🇷 A nuclear submarine is sitting off Cyprus loaded with 20 nuclear-armed missiles while Iran and Israel trade fire. Retired Colonel Anthony Aguilar just walked through exactly how fast this spirals out of anyone's control. "We are now in another level of escalatory territory that is out of Donald Trump's control." The off-ramp isn't just narrow. It may not exist. @AnthonyAgu88102

Mario Nawfal

129,077 görüntüleme • 2 ay önce

Last semester, I taught Reinforcement Learning class again at UCLA. Together with my amazing TAs Matthew and Caiyuan, we built a mini-project: MetaDrive Arena 🚗🤖 Students applied what they learned in class, trained RL agents, and competed on a live leaderboard. The results were incredible, with 94 agents, 2K submissions, and 130K matches. We saw tons of creativity, clever ideas, and real progress in learning. We’re now releasing it publicly to support RL education and experimentation. Try it out and train your own agent at 🔗

Last semester, I taught Reinforcement Learning class again at UCLA. Together with my amazing TAs Matthew and Caiyuan, we built a mini-project: MetaDrive Arena 🚗🤖 Students applied what they learned in class, trained RL agents, and competed on a live leaderboard. The results were incredible, with 94 agents, 2K submissions, and 130K matches. We saw tons of creativity, clever ideas, and real progress in learning. We’re now releasing it publicly to support RL education and experimentation. Try it out and train your own agent at 🔗

Bolei Zhou

25,253 görüntüleme • 4 ay önce

Opus 4.6 vs GPT-5.4 (High) (6/9) Testing out how well these models to with cloth physics simulation, and honestly both where pretty close here but my god I am shocked how good GPT 5.4 is with optimizing this scene whilst opus 4.6 felt quite janky and lower fps. GPT 5.4 also got the cutting working properly whilst Opus was just floating about. I do want to heavily stress the fact that making this prompt, Opus 4.6 cost the whole 100% of the session ($20 pro plan) + $3 in credits whilst GPT 5.4 only cost 10% of the usage. prompt: Build a premium Vite + React interactive cloth simulation of a cozy room curtain using real particle/constraint physics, with structural + shear springs, damping, collisions, and stable real-time deformation. Users should be able to drag the curtain, add wind, and cut constraints, while the experience stays polished, visually elegant, and numerically stable.

Opus 4.6 vs GPT-5.4 (High) (6/9) Testing out how well these models to with cloth physics simulation, and honestly both where pretty close here but my god I am shocked how good GPT 5.4 is with optimizing this scene whilst opus 4.6 felt quite janky and lower fps. GPT 5.4 also got the cutting working properly whilst Opus was just floating about. I do want to heavily stress the fact that making this prompt, Opus 4.6 cost the whole 100% of the session ($20 pro plan) + $3 in credits whilst GPT 5.4 only cost 10% of the usage. prompt: Build a premium Vite + React interactive cloth simulation of a cozy room curtain using real particle/constraint physics, with structural + shear springs, damping, collisions, and stable real-time deformation. Users should be able to drag the curtain, add wind, and cut constraints, while the experience stays polished, visually elegant, and numerically stable.

Dev Ed

414,465 görüntüleme • 5 ay önce