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。

Introducing ScienceBuddy — a free workspace for scientific agents that improve through researcher collaboration. Use GPT-6 in ScienceBuddy at no cost. GPU-accelerated, and fused with the JEV framework. 🧵 Two loops: 🔹 Inner loop — refines the agent harness 🔹 Outer loop — trains the model with rubric-guided RL... show more

PhAI Labs

3,801 subscribers

1,261,542 次观看 • 1 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

35 条评论

liam. 的头像

liam.1 个月前

congrats team!

Franklin 的头像

Franklin1 个月前

The ambition is splendid. I’d simply ask you to put the before-and-after results beside the impressive vocabulary. Preferably in the same size type. You have my attention, but if it’s learning to earn praise instead of finding the truth, you’ve built a very expensive suck-up.

Justin Zhou 的头像

Justin Zhou1 个月前

congrats the launch. love the science agent. will try it

叫我阿杭的头像

叫我阿杭1 个月前

确实有东西啊，我觉得这个模型拿来做自媒体的调研，真太牛了

我真的没有拼多多的头像

我真的没有拼多多1 个月前

非常好用！我让它调研了一下睡眠跟注意力之间的关系

Ella Tech & Tool 的头像

Ella Tech & Tool1 个月前

Free GPT-6 GPU RL loops? This is huge

知识猫AI实验室的头像

知识猫AI实验室1 个月前

关键是GPT 6 免费用，太良心了吧

Connect Despirit 的头像

Connect Despirit1 个月前

Have you tried GPT-6 in ScienceBuddy yet

EyeingAI 的头像

EyeingAI1 个月前

Researchers are gonna have fun with this one.

Aria Tech 的头像

Aria Tech1 个月前

ScienceBuddy looks really impressive recursive improvement idea is super exciting for research

Jeremy Bosma 的头像

Jeremy Bosma1 个月前

Separating harness and model loops is useful

雪踏乌云的头像

雪踏乌云1 个月前

试了下，确实不错，证据源列出的很清晰

摸鱼巨匠🔨 的头像

摸鱼巨匠🔨1 个月前

非常好，我已经推荐给我很多博士生同学了

Sadok 的头像

Sadok1 个月前

getting free gpt-6 access for science is a huge deal

来碗牛肉粉的头像

来碗牛肉粉1 个月前

朋友安利的，自己体验了下真的非常强！科研界真正的WorkBuddy

Aaliya 的头像

Aaliya1 个月前

quite impressive combining researcher collaboration with rubric guided RL could give scientific agents a more structured way to improve over time.

阿良｜AI 工作流的头像

阿良｜AI 工作流1 个月前

把科学家的反馈变成代理进化燃料，方向对，双循环自我改进能否真发现新科学，还得看实验。

Rachel🥥 的头像

Rachel🥥1 个月前

很好用，科研人员必备

Elara AI 的头像

Elara AI1 个月前

Recursive-in-Recursive self-improvement with inner loop for harness and outer loop for model training is a really elegant approach - free access with GPT-6 makes it easy to test too

Kirill 的头像

Kirill1 个月前

That's interesting! Thanks, I'll check it out right away.

Evia AI 的头像

Evia AI1 个月前

Recursive-in-Recursive Self-Improvement with inner loop for harness + outer loop with rubric-guided RL is a fascinating approach. Free access to GPT-6 + GPU acceleration + JEV framework for scientific agents is huge for researchers. Excited to see what the community builds at

QuietNode 的头像

QuietNode1 个月前

sciencebuddy goes hard 🔥

Zico 的头像

Zico1 个月前

sciencebuddy lfg 🚀

Yingcheng Charles Wu 的头像

Yingcheng Charles Wu1 个月前

🔬Try it free:

Shraddha Bharuka 的头像

Shraddha Bharuka1 个月前

Really interesting approach to improving scientific AI agents.

程序员鱼皮的头像

程序员鱼皮1 个月前

已经用上了，用来研究新东西很不错 😉

SpreadX AI 的头像

SpreadX AI1 个月前

Free science agent! Thanks team

The Calm Warrior 的头像

The Calm Warrior1 个月前

GPT-6 for free? lol what is JEV though

JuanAI 的头像

JuanAI1 个月前

Good step!

marcus 的头像

marcus1 个月前

banger launch team

Ravindar Bishnoi Mukam Nokha 的头像

Ravindar Bishnoi Mukam Nokha1 个月前

GPT-6 at no cost? wait what lol

gigaFlip.eth 的头像

gigaFlip.eth1 个月前

GPU-accelerated JEV setup works out of the box?

Cameron Ng 的头像

Cameron Ng1 个月前

Rubric guided RL for scientific harness is a neat approach

ShadowAguy 的头像

ShadowAguy1 个月前

The inner/outer loop structure is the detail that actually matters. Most agent frameworks just stack tools on tools without a refinement mechanism. Does ScienceBuddy use the outer loop to benchmark against other researchers or just self-improve?

crypt0messenger 的头像

crypt0messenger1 个月前

What kind of rubrics are you using for the scientific evaluation in the outer loop?

相关视频

Asked one question in plain English and got back real ADMET properties with the values laid out. No code. No local setup. No install. ScienceBuddy is a free research workspace from PhAI Labs. → Ask in plain language → It plans, retrieves and runs the analysis → Every step is inspectable → Result files download 224 scientific tools integrated across genomics, pharmacology and bioimaging. Runs in the browser at ScienceBuddy is currently a Preview release. Verify results yourself before using them in research. PhAI Labs #ScienceBuddy #PhAILabs #AIforScience

Asked one question in plain English and got back real ADMET properties with the values laid out. No code. No local setup. No install. ScienceBuddy is a free research workspace from PhAI Labs. → Ask in plain language → It plans, retrieves and runs the analysis → Every step is inspectable → Result files download 224 scientific tools integrated across genomics, pharmacology and bioimaging. Runs in the browser at ScienceBuddy is currently a Preview release. Verify results yourself before using them in research. PhAI Labs #ScienceBuddy #PhAILabs #AIforScience

Dean W. Perkins

28,309 次观看 • 1 个月前

I just tried ScienceBuddy, a free AI research workspace from PhAI Labs. I uploaded research documents and asked it to analyze the evidence and build a structured table. What stood out: → Plans the research task → Retrieves relevant sources → Runs research tools and analysis → Shows the steps in the Trajectory panel → Gives you a structured final result You can actually inspect what happened instead of just getting a final answer. Try it: Research paper: Technical report: GitHub: ScienceBuddy is currently a Preview release. Please verify results yourself before using them in research. #ScienceBuddy #PhAILabs #AIforScience

I just tried ScienceBuddy, a free AI research workspace from PhAI Labs. I uploaded research documents and asked it to analyze the evidence and build a structured table. What stood out: → Plans the research task → Retrieves relevant sources → Runs research tools and analysis → Shows the steps in the Trajectory panel → Gives you a structured final result You can actually inspect what happened instead of just getting a final answer. Try it: Research paper: Technical report: GitHub: ScienceBuddy is currently a Preview release. Please verify results yourself before using them in research. #ScienceBuddy #PhAILabs #AIforScience

Aryan Rakib

63,461 次观看 • 1 个月前

Tried ScienceBuddy, a free AI research workspace from PhAI Labs. Asked it in plain English to look up the ADMET properties of ibuprofen and report key values. It planned the work, queried PubChem and PubMed, and returned absorption, protein binding, CYP2C9 metabolism, half-life and clearance, with sources for each value and notes on where the literature disagrees. The part I liked: the Trajectory tab. Every tool call and step is shown in order, so you can see exactly what it searched instead of taking the answer on faith. Result files can also be downloaded for your own checking. It integrates 224 scientific tools across genomics, molecular and cancer biology, pharmacology, bioimaging and literature retrieval. Web-based, no local setup, no code. Free to try: ScienceBuddy is currently a Preview release. Please verify results yourself before using them in research. #ScienceBuddy #PhAILabs #AIforScience

Tried ScienceBuddy, a free AI research workspace from PhAI Labs. Asked it in plain English to look up the ADMET properties of ibuprofen and report key values. It planned the work, queried PubChem and PubMed, and returned absorption, protein binding, CYP2C9 metabolism, half-life and clearance, with sources for each value and notes on where the literature disagrees. The part I liked: the Trajectory tab. Every tool call and step is shown in order, so you can see exactly what it searched instead of taking the answer on faith. Result files can also be downloaded for your own checking. It integrates 224 scientific tools across genomics, molecular and cancer biology, pharmacology, bioimaging and literature retrieval. Web-based, no local setup, no code. Free to try: ScienceBuddy is currently a Preview release. Please verify results yourself before using them in research. #ScienceBuddy #PhAILabs #AIforScience

Shraddha Bharuka

68,505 次观看 • 1 个月前

Ex-Google engineer explained AI agent loops, harness, evals in 20 minutes - better than 500$ courses. trace every run → judge it with an LLM → diagnose → fix → ship. That loop is how agents self-improve over time. Agent loops + memory + harness + evals - thats the stack. Watch it, then save the framework below.

Ex-Google engineer explained AI agent loops, harness, evals in 20 minutes - better than 500$ courses. trace every run → judge it with an LLM → diagnose → fix → ship. That loop is how agents self-improve over time. Agent loops + memory + harness + evals - thats the stack. Watch it, then save the framework below.

Movez

568,462 次观看 • 4 个月前

Most AI research tools give you an answer and hide how they got there. This one shows the whole thing. Asked ScienceBuddy to plan an RNA-seq differential expression workflow and the Trajectory panel laid out every step in order. → The plan it wrote → Each tool call → The actual code it ran → The observations at every stage Nothing hidden, everything checkable. Free, from PhAI Labs. ScienceBuddy is currently a Preview release. Verify results yourself before using them in research. PhAI Labs #ScienceBuddy #PhAILabs #AIforScience

Most AI research tools give you an answer and hide how they got there. This one shows the whole thing. Asked ScienceBuddy to plan an RNA-seq differential expression workflow and the Trajectory panel laid out every step in order. → The plan it wrote → Each tool call → The actual code it ran → The observations at every stage Nothing hidden, everything checkable. Free, from PhAI Labs. ScienceBuddy is currently a Preview release. Verify results yourself before using them in research. PhAI Labs #ScienceBuddy #PhAILabs #AIforScience

liam.

27,789 次观看 • 1 个月前

Anthropic Research Lead: "99% of our engineers run swarms of 300+ self-improving agents Close the loop and give the model a way to verify its own output" In a 20-minute session, an Anthropic team member walks through how to build agents that improve after every run Agents → Harness → Loops → Self-Improvement the real setup is Claude running through systems that check, correct, and upgrade themselves the part most agent courses skip this is better than most $1,300 agent engineering courses bookmark and watch the talk then read the full article below

Anthropic Research Lead: "99% of our engineers run swarms of 300+ self-improving agents Close the loop and give the model a way to verify its own output" In a 20-minute session, an Anthropic team member walks through how to build agents that improve after every run Agents → Harness → Loops → Self-Improvement the real setup is Claude running through systems that check, correct, and upgrade themselves the part most agent courses skip this is better than most $1,300 agent engineering courses bookmark and watch the talk then read the full article below

Callan

14,956 次观看 • 1 个月前

The hardest part of a literature review was never the reading. It was comparing four papers without losing track of which one said what. Dropped four open-access PDFs into ScienceBuddy and asked what evidence they gave on immunotherapy resistance. → No code, no local setup → Plain language question → Every claim traced back to its paper → Table downloads straight out Free, from PhAI Labs, at Biomedicine only for now, and sign-up needs a .edu address at the moment. ScienceBuddy is currently a Preview release. Verify results yourself before using them in research. PhAI Labs #ScienceBuddy #PhAILabs #AIforScience

The hardest part of a literature review was never the reading. It was comparing four papers without losing track of which one said what. Dropped four open-access PDFs into ScienceBuddy and asked what evidence they gave on immunotherapy resistance. → No code, no local setup → Plain language question → Every claim traced back to its paper → Table downloads straight out Free, from PhAI Labs, at Biomedicine only for now, and sign-up needs a .edu address at the moment. ScienceBuddy is currently a Preview release. Verify results yourself before using them in research. PhAI Labs #ScienceBuddy #PhAILabs #AIforScience

marcus

50,657 次观看 • 1 个月前

Build reliable agents for long-running tasks. Loop Engineering with LangChain Deep Agents: add loops that check output, trigger runs on events, and improve the harness from traces across runs. Taught by Sydney Runkle and Esmaeil Gargari. Enroll for free:

Build reliable agents for long-running tasks. Loop Engineering with LangChain Deep Agents: add loops that check output, trigger runs on events, and improve the harness from traces across runs. Taught by Sydney Runkle and Esmaeil Gargari. Enroll for free:

DeepLearning.AI

38,533 次观看 • 14 天前

Introducing Ricursive Intelligence, a frontier AI lab enabling a recursive self-improvement loop between AI and the chips that fuel it. Learn more at

Introducing Ricursive Intelligence, a frontier AI lab enabling a recursive self-improvement loop between AI and the chips that fuel it. Learn more at

Ricursive Intelligence

534,090 次观看 • 11 个月前

Ex-Google engineer reveals how to build AI agent loops, harnesses, LLM ops, and evals in 19 minutes. Trace → evaluate → diagnose → fix → ship → repeat. That loop is how agents self-improve over time. Agentic loops + harness + memory + evals - that’s the senior engineer stack. This is better than $500 paid courses on the same topic, explained in under 20 minutes. Watch it, then save the framework below.

Ex-Google engineer reveals how to build AI agent loops, harnesses, LLM ops, and evals in 19 minutes. Trace → evaluate → diagnose → fix → ship → repeat. That loop is how agents self-improve over time. Agentic loops + harness + memory + evals - that’s the senior engineer stack. This is better than $500 paid courses on the same topic, explained in under 20 minutes. Watch it, then save the framework below.

Codez

174,272 次观看 • 4 个月前

Today we’re introducing PROWL-2, where agents and their world model improve through recursive learning. Agents expose errors in imagination, and repairing those errors enables further learning. We believe this open-ended learning is a critical step towards superintelligence.

Today we’re introducing PROWL-2, where agents and their world model improve through recursive learning. Agents expose errors in imagination, and repairing those errors enables further learning. We believe this open-ended learning is a critical step towards superintelligence.

Odyssey

39,032 次观看 • 1 个月前

Introducing self-improvement loops. The concept is simple: what if agents could improve Skills by reviewing past conversations? Here's the three step loop: - Score conversations from criteria you define - Isolate failures - Generate skill improvements

Introducing self-improvement loops. The concept is simple: what if agents could improve Skills by reviewing past conversations? Here's the three step loop: - Score conversations from criteria you define - Isolate failures - Generate skill improvements

Warp

128,782 次观看 • 2 个月前

Can an agent explore a new environment, learn its causal structure, and keep improving without updating its model weights? We introduce RSIAgent, a framework for recursive self-improvement through autonomous exploration. Using Kimi-K3 and GLM-5.3 as base models, RSIAgent outperforms GPT-6 Astra on both OSWorld 2.0 and Agents’ Last Exam. RSIAgent decides what to explore, executes tasks, verifies outcomes, and consolidates stable action-condition-outcome relationships into memory for future use. With the underlying model weights fixed, RSIAgent achieves: - 78.98% Partial Score on OSWorld 2.0 (0808 offline), compared with 72.60% for GPT-6 Astra - 84.82% on Agents’ Last Exam (Near-term), compared with 82.26% for GPT-6 Astra We call this Scaling Experience. Agents can continue improving by acquiring, verifying, and reusing their own experience while the underlying model weights remain fixed. Links in the reply below.

Can an agent explore a new environment, learn its causal structure, and keep improving without updating its model weights? We introduce RSIAgent, a framework for recursive self-improvement through autonomous exploration. Using Kimi-K3 and GLM-5.3 as base models, RSIAgent outperforms GPT-6 Astra on both OSWorld 2.0 and Agents’ Last Exam. RSIAgent decides what to explore, executes tasks, verifies outcomes, and consolidates stable action-condition-outcome relationships into memory for future use. With the underlying model weights fixed, RSIAgent achieves: - 78.98% Partial Score on OSWorld 2.0 (0808 offline), compared with 72.60% for GPT-6 Astra - 84.82% on Agents’ Last Exam (Near-term), compared with 82.26% for GPT-6 Astra We call this Scaling Experience. Agents can continue improving by acquiring, verifying, and reusing their own experience while the underlying model weights remain fixed. Links in the reply below.

Biwei Huang

63,271 次观看 • 1 个月前

spent hours doing literature reviews by hand before i tried ScienceBuddy it's a free AI research workspace from PhAI Labs, no payment or waitlist, just an .edu email required right now → uploaded [3-5 papers on your real topic] and asked it to build an evidence table with sources → Trajectory panel showed the whole plan, every tool call and the code, nothing was a black box → downloaded the [table / plot] straight from the Results panel → still had to check the output myself before trusting it for real work it's a Preview release, verify anything it gives you before using it in actual research the [evidence table / plan] step actually saved time, the rest i'd still do by hand #ScienceBuddy #PhAILabs #AIforScience

spent hours doing literature reviews by hand before i tried ScienceBuddy it's a free AI research workspace from PhAI Labs, no payment or waitlist, just an .edu email required right now → uploaded [3-5 papers on your real topic] and asked it to build an evidence table with sources → Trajectory panel showed the whole plan, every tool call and the code, nothing was a black box → downloaded the [table / plot] straight from the Results panel → still had to check the output myself before trusting it for real work it's a Preview release, verify anything it gives you before using it in actual research the [evidence table / plan] step actually saved time, the rest i'd still do by hand #ScienceBuddy #PhAILabs #AIforScience

kuch.

26,681 次观看 • 1 个月前

About a year ago, we launched the Agent-to-Agent protocol because AI agents shouldn’t work in silos. Today, developers are using A2A to coordinate specialized agents across enterprise data systems, commerce workflows, DevOps environments, and scientific research. FoldRun is one example, simplifying petabyte-scale protein modeling through agent-to-agent collaboration. Learn how to build with it in the blog:

About a year ago, we launched the Agent-to-Agent protocol because AI agents shouldn’t work in silos. Today, developers are using A2A to coordinate specialized agents across enterprise data systems, commerce workflows, DevOps environments, and scientific research. FoldRun is one example, simplifying petabyte-scale protein modeling through agent-to-agent collaboration. Learn how to build with it in the blog:

Google for Developers

35,497 次观看 • 4 个月前

Introducing LangSmith Fleet. Agents for every team. → Build agents with natural language → Share and control who can edit, run, or clone each agent → Manage authentication with agent identity → Approve actions with human-in-the-loop → Track and audit actions with tracing in LangSmith Observability Try Fleet:

Introducing LangSmith Fleet. Agents for every team. → Build agents with natural language → Share and control who can edit, run, or clone each agent → Manage authentication with agent identity → Approve actions with human-in-the-loop → Track and audit actions with tracing in LangSmith Observability Try Fleet:

LangChain

9,398,215 次观看 • 7 个月前

🤝 AI Agents That Work Together — Collaboration Made Simple! In this demo, John explores one of Agent Forge’s most powerful features: agent-to-agent collaboration. With Agent Forge, you can create agents with different roles, connect them, and watch them collaborate to solve complex tasks. Modular, scalable, and fully automated. 🔗 Try here:

🤝 AI Agents That Work Together — Collaboration Made Simple! In this demo, John explores one of Agent Forge’s most powerful features: agent-to-agent collaboration. With Agent Forge, you can create agents with different roles, connect them, and watch them collaborate to solve complex tasks. Modular, scalable, and fully automated. 🔗 Try here:

AITECH CLOUD NETWORK

43,212 次观看 • 1 年前

Introducing Collaborative Reasoner: a framework to improve collaborative reasoning in language models. Collaborative Reasoner paves the way for developing social agents that can partner with humans and other agents. Read the research paper and download the code.

Introducing Collaborative Reasoner: a framework to improve collaborative reasoning in language models. Collaborative Reasoner paves the way for developing social agents that can partner with humans and other agents. Read the research paper and download the code.

AI at Meta

58,590 次观看 • 1 年前

Our latest LangChain Academy course – Building Ambient Agents with LangGraph – is now available! Most agents today handle one request at a time through chat interfaces. But as models have improved, agents can now run in the background – and take on long-running, complex tasks. LangGraph is built for these “ambient agents,” with support for human-in-the-loop workflows and memory. LangGraph Platform provides the infrastructure to run these agents at scale, and LangSmith helps you observe, evaluate, and improve your agents. Together, they make it easier to build reliable, production-ready agents. In this course, you’ll: 📩 Learn the fundamentals of LangGraph and how to build your own email agent 📊 Evaluate your agent using LangSmith ✅ Add human-in-the-loop for reviews, and memory so your agent adapts over time 🚀 Deploy your agent with LangGraph Platform and connect it to Gmail

Our latest LangChain Academy course – Building Ambient Agents with LangGraph – is now available! Most agents today handle one request at a time through chat interfaces. But as models have improved, agents can now run in the background – and take on long-running, complex tasks. LangGraph is built for these “ambient agents,” with support for human-in-the-loop workflows and memory. LangGraph Platform provides the infrastructure to run these agents at scale, and LangSmith helps you observe, evaluate, and improve your agents. Together, they make it easier to build reliable, production-ready agents. In this course, you’ll: 📩 Learn the fundamentals of LangGraph and how to build your own email agent 📊 Evaluate your agent using LangSmith ✅ Add human-in-the-loop for reviews, and memory so your agent adapts over time 🚀 Deploy your agent with LangGraph Platform and connect it to Gmail

LangChain

33,192 次观看 • 1 年前

Anthropic just released 5 Claude workshops on the new skill every AI builder needs: Loop Engineering: 00:00 - Stop babysitting your agents 37:07 - Build a proactive Claude Code loop 1:11:23 - Build a production-ready Managed Agent 1:38:46 - Teach agents to learn from your team 2:07:15 - Hill-climb your agent with eval loops It is a 2 hour and 25 minute blueprint for Loop Engineering: proactive workflows, managed agents, team feedback, evals, and self-improvement. Watch it today, and read the article below to improve your skills with Fable 5.

Anthropic just released 5 Claude workshops on the new skill every AI builder needs: Loop Engineering: 00:00 - Stop babysitting your agents 37:07 - Build a proactive Claude Code loop 1:11:23 - Build a production-ready Managed Agent 1:38:46 - Teach agents to learn from your team 2:07:15 - Hill-climb your agent with eval loops It is a 2 hour and 25 minute blueprint for Loop Engineering: proactive workflows, managed agents, team feedback, evals, and self-improvement. Watch it today, and read the article below to improve your skills with Fable 5.

Alex

47,626 次观看 • 4 个月前