Flows Agent lets you iterate through conversation. Tell it to try a warmer voice, swap the background, or generate a version in Spanish. The agent modifies the pipeline and re-runs without rebuilding from scratch.
SYSTEMS
-
Clippy is back, powered by new Microsoft MAI models for code judgment
By
–
CLIPPY 👏 IS 👏 BACK 👏
— Charly Wargnier (@DataChaz) 4 juin 2026
but this time he’s powered by frontier AI models ready to judge your code 🙈
Microsoft’s new MAI models just dropped on @aimlapi
They recreated Windows XP using MAI-Thinking-1 + @crewAIInc, and brought our fave assistant to life using MAI-Image 2.5 👀↓ https://t.co/AkCibYJJFyCLIPPY IS BACK but this time he’s powered by frontier AI models ready to judge your code Microsoft’s new MAI models just dropped on @aimlapi They recreated Windows XP using MAI-Thinking-1 + @crewAIInc
, and brought our fave assistant to life using MAI-Image 2.5 ↓ -
Full study on efficient verifiers for legal agents
By
–
Read our full study: https://
langchain.com/blog/designing
-efficient-verifiers-for-legal-agents
…? -
Easy AI demos, hard production: weaknesses exposed
By
–
Here is the uncomfortable reality I observe in companies: AI demos are easy. AI in production is not. Once AI moves past the pilot stage, it begins to expose every weakness in the foundation: → Fragmented systems
→ Applications -
AI pilots fail mainly because of the company, not the model
By
–
Most AI pilots do not fail because the model is weak.
— Ronald van Loon (@Ronald_vanLoon) 4 juin 2026
They fail because the enterprise underneath it was never built for production AI.
→ Data volume
→ Latency
→ Deployment cycles
→ Legacy dependencies
→ Technical debt
This is the infrastructure problem nobody is… pic.twitter.com/pVbkLn2f1uMost AI pilots do not fail because the model is weak. They fail because the underlying company was never designed for production AI. → Data volume
→ Latency
→ Deployment cycles
→ Legacy dependencies
→ Technical debt It's -
Gautam Kamath thanks Peter for collaborative Byzantine robustness work
By
–
Thanks Peter! Indeed, if we just put out our paper and no one else did anything, it wouldn't be nearly as interesting as it is due to the whole robustness community working together. As I recall, you famously also worked on this area (Byzantine robustness)
-
AEP-001 GoalOS Proof-of-Evolution Constitution Standard
By
–
New standard for the agent era: AEP-001 — GoalOS Proof-of-Evolution Constitution Commit → Execute → Prove → Evolve. No proof, no evolution.
No eval, no propagation.
No rollback, no release. This is Proof-Carrying Intelligence. https://
montrealai.github.io/proof-gradient
/standards/AEP-001/
… #GoalOS -
Pipeline order is a hyperparameter for optimizing LLM execution strategies
By
–
5/5 Takeaway: pipeline order is a hyperparameter. If you're already paying for parallel rollouts, reuse them – they're relevant context, not just candidate answers. Full write-up: [
https://
ai21.com/blog/first-sca
le-then-enrich-how-the-right-execution-strategy-helped-us-reach-state-of-the-art-on-swe-rebench/?utm_source=org-twitter
…] -
AI21 Labs: ReAct Agent Performance with Enrichment and Scaling Strategies
By
–
2/5 Started with a baseline: classic ReAct agent (GPT-5.2), single Docker-terminal tool. Baselines on the slice: vanilla 53.8%, enrich-only 55.6%, scale-only (n=5 + LLM judge) 55.4%, enrich-then-scale 57.7%.
-
AI21 Labs: Reversing agent pipeline order achieves SOTA results
By
–
1/5 Our latest Labs in Front piece: Agent pipeline order matters. By reversing a common agent recipe – scale first, enrich second – we reached SOTA on a Dec ‘25 to Mar ‘26 slice (123 issues): 60.9%.