Shaiann Collins

Hey friend(: ☆ BSHS ☆ Sophomore

[Shaiann Collins @Miss_ShaiC]

@42space 13 out of 78 is wild stats ngl

[Ling Yang @LingYang_PU]

Can research agents get better at science by learning from the research they do?

Introducing DiscoveryRSI 🧬: a framework that turns hypothesis-driven research into a recursive self-improvement loop.

📄 Paper: https://t.co/aBpR5GcChg
💻 Code: https://t.co/XjUmq6xt6a
🔴 Live Arena: https://t.co/Z4F6k8fqDT

🔹 What is a valuable hypothesis?
We define it as a joint object H = (claim, investigation plan, decision criteria, alternative explanations). A hypothesis is valuable if resolving it would change a research decision. It is testable if the plan can produce evidence that supports, weakens, or bounds the claim.

🔹 How it works
- Lead, Generation, Evolution, Critique and Progress agents grow hypothesis trees through evolve / repair / crossover / restart
- Every version stays linked to its evidence, data, assessments and contribution records
- Selected plans compile into executable Harbor tasks

🔹 Three self-improvement mechanisms
① Hypothesis credit optimization: blinded TrueSkill assessments assign bounded credit to the agent steps that actually contributed, and DiscoveryRSI-27B is trained only on those steps (Qwen3.8-27B init, 256K context; 2,397 of 15,792 trajectories kept; 2.12% of tokens supervised)
② Harness replay: diagnose recurring failures such as repeated exploration, failed tool use, lost artifacts and premature closure, then refine the workflow
③ Rediscovery: run the accepted model and harness updates on fresh questions

🔹 Results (DFM-Bench, 4 life-science domains)
- Hypothesis quality 4.86 → 5.18 → 5.39 over two rounds
- Falsifiability 4.00 → 6.20, alternative explanations 4.44 → 5.98, data resolvability 3.73 → 4.99
- Evolve/repair share 77.5% → 91.7%, restarts 21.7% → 5.0%: the agent learns to keep developing an idea instead of starting over
- Caveat: importance and novelty dip slightly. The model learns to design better tests, not yet to choose bolder questions.

🔹 Expert review
27 physicians and bioscience researchers scored 292 anonymized proposals:
- 41% of DiscoveryRSI-27B revisions improve quality vs 31% (GPT-5.6 Sol) and 25% (GPT-6 Luna)
- Only 12% make things worse vs 32% and 31%

🔹 Does it anticipate real science?
Against 16,440 publications filtered by each model's cutoff, 6.3% of DiscoveryRSI-27B's hypotheses connect to later published research (16.9% in therapeutic targets). Comparison: DeepSeek 5.4%, GLM 2.8%, GPT 2.5%, Claude 2.3%.

Watch six models do research in real time in DiscoveryRSI Arena, and follow every proposal, critique and compile step 👇
https://t.co/Z4F6k8fqDT

Co-led by Jianyuan Zhong, Chenwei Wu & Zhaochen Yu, with wonderful collaborators across PhAI Labs, CUHK, Stanford, MIT, Harvard Medical School, Oxford, Princeton, PKU and more.

Grateful to @james_y_zou @Jiaru_Zou @Charles_Y_Wu @zhenfei_yin_yzf 🙏

[Shaiann Collins @Miss_ShaiC]

@0xbeepit most of it is just infinite loops burning compute tbh

[Shaiann Collins @Miss_ShaiC]

@abbyxuu Ok the non-redundant factor prediction idea actually makes sense

[Shaiann Collins @Miss_ShaiC]

@Killa1409 @Teamily_AI Zoe and yenn made that look way too easy

[Shaiann Collins @Miss_ShaiC]

@meiwoo_ @Teamily_AI most agent frameworks ignore human-in-the-loop UX, so glad @Teamily_AI made it 1-click

[Shaiann Collins @Miss_ShaiC]

@mwixamwixa2 @simular Claude mcp integrations getting absurdly fast to deploy

[Shaiann Collins @Miss_ShaiC]

@RogerioPontes1 What happens when the win desktop app forces an unannounced modal update though?

[Shaiann Collins @Miss_ShaiC]

@ramanrope osworld 2.0 specs look wild

[Shaiann Collins @Miss_ShaiC]

@NAZEERNAZNI Robosecretary is a great term for this

[Shaiann Collins @Miss_ShaiC]

@khushiirl lfg teamily 🚀

[Shaiann Collins @Miss_ShaiC]

@dee_naliaks @Teamily_AI dungaree button exhibit is gold

[Teamily AI @Teamily_AI]

150,000 small business owners and employees are already using Teamily AI. Today, we're opening it to everyone: your personal AI agent and an entire team of AI agents, ready to work for you and your business.

Meet Teamily AI 2.0 — the human–agent platform where you and always-on AI agents get work done together in one unified, context-aware space.

We've taken four major capabilities to the next level to help businesses build and grow faster:

1. Real-Time Human–Agent Collaboration. Humans and AI agents work together across chats, meetings, studios, and communities to research, analyze, and create websites, dashboards, videos, slides, documents, and more.

2. Unified Context & Living Memory. One connected context across conversations, meetings, channels, and workspaces. Your AI agents remember, learn, and understand you over time—delivering smarter, more personalized results.

3. Custom Agents & Agent Teams. Create and orchestrate AI agent teams in one click. Manage agents through Agent Workspace, connect them to Telegram, WhatsApp, iMessage, and Slack, or integrate them into your products via Agent API.

4. Intelligent Multi-Model Routing. Our proprietary models work alongside leading models from OpenAI, Anthropic, Gemini, DeepSeek, Qwen, Kimi, MiniMax, and more—automatically routing each task to the best model across text, image, video, and voice.

Teamily AI 2.0 is built for every business owner, every employee, and every customer.

America has 36 million small businesses. 98% have fewer than 20 employees. Many are teams of one. At Teamily, we believe AI should democratize opportunity, giving everyone the power to achieve more.

You have the business. Now you have the team.

Try Teamily AI 2.0 for free → https://t.co/75i5LmS8OV

[Shaiann Collins @Miss_ShaiC]

a daily trends loop that splits tech for zoe and social for yenn

Teamily AI is a Humans + Agents platform built for this kind of Team AI workflow.
re-run the report with one click at https://t.co/FHZGpukPN5

[Shaiann Collins @Miss_ShaiC]

@Charles_Y_Wu Shoutout salesforce for helping out with student registration

[Shaiann Collins @Miss_ShaiC]

@AngryTomtweets Thought local AI was a gimmick until seeing local memory done right

[Shaiann Collins @Miss_ShaiC]

@unycornwarria 27B params sitting right on the desk is wild

[Shaiann Collins @Miss_ShaiC]

Meet the Personal AI Computer designed without a monthly subscription. Oki Home serves as a Memory Computer that runs local AI on hardware in your house. Details at https://t.co/jBZOgyY3BR.

[Shaiann Collins @Miss_ShaiC]

@hellotradeapp Spot AND perps from one balance? huge if true