After two days with Claude Fable 5, the best way I can describe it is 'relentlessly proactive' – here is an example where I inserted a screenshot of a bug and it started custom Python CORS servers and used pyobjc-framework-Quartz to capture
AGENTS
-
Agentic AI orchestrates EC workflows between design tools
By
–
EC workflows are increasingly defined by the connections between tools, not the tools themselves.
— NVIDIA (@nvidia) 11 juin 2026
Agentic AI has the potential to streamline everything from early design exploration to visualization by orchestrating work across applications, reducing friction and accelerating… https://t.co/Bf0KHsEBToEC workflows are increasingly defined by the connections between tools, rather than the tools themselves. Agentic AI has the potential to streamline everything from early design exploration to visualization, by orchestrating work across the
-
Loop engineering automates agent prompting loops for efficient task completion
By
–
Loop engineering is everywhere right now.
— Louis-François Bouchard 🎥🤖 (@Whats_AI) 11 juin 2026
Instead of you prompting the agent, checking what broke, and telling it to go again, you build a small system that runs that loop for you. It keeps the agent going until the work is actually done.
I explain it all in this video 👇 pic.twitter.com/EUuULCDoCOLoop engineering is everywhere right now. Instead of you prompting the agent, checking what broke, and telling it to go again, you build a small system that runs that loop for you. It keeps the agent going until the work is actually done. I explain it all in this video
-
Show, Don’t Tell: Provide References for Better Results
By
–
3. Show, Don't Tell
— Replit ⠕ (@Replit) 11 juin 2026
Give Agent something to reference: screenshots, website links, or files. Want it to match a design? Drop in the mockup. The more it has to look at, the closer it gets to what you actually want. pic.twitter.com/g9CosRTvhO3. Show, Don't Tell Give Agent something to reference: screenshots, website links, or files. Want it to match a design? Drop in the mockup. The more it has to look at, the closer it gets to what you actually want.
-
How to prompt like a pro with Replit Agent
By
–
How to prompt like a pro with Replit Vague prompts just mean more rewrites. Here's how to get Agent to build the right thing the first time. Open thread ↓
-
Teach Replit Agent your conventions with Custom Instructions and Skills
By
–
AI agents are powerful, but they don’t remember your preferences.
— Replit ⠕ (@Replit) 11 juin 2026
So you end up repeating instructions- How you structure projects. Your brand guidelines.
You can now teach Replit Agent your conventions with Custom Instructions and Skills.
It'll take them into account for… pic.twitter.com/WntiVxyzBOAI agents are powerful, but they don’t remember your preferences. So you end up repeating instructions- How you structure projects. Your brand guidelines. You can now teach Replit Agent your conventions with Custom Instructions and Skills. It'll take them into account for
-
DARPA AI Cyber Challenge: multi-agent LLM discovers zero-day vulnerabilities
By
–
At the DARPA AI Cyber Challenge, Team Atlanta from Georgia Institute of Technology demonstrated a multi-agent LLM framework that discovered zero-day vulnerabilities in large codebases and automatically generated, tested, and deployed functional patches without human intervention.… pic.twitter.com/gBxsiw7Jgl
— Lucian Fogoros (@fogoros) 11 juin 2026At the DARPA AI Cyber Challenge, Team Atlanta from Georgia Institute of Technology demonstrated a multi-agent LLM framework that discovered zero-day vulnerabilities in large codebases and automatically generated, tested, and deployed functional patches without human intervention.
-
Perplexity Computer adds Deep Research as native skill
By
–
Perplexity Computer is an agent harness that just keeps delivering. Deep Research is now a native skill inside Computer (you don’t have to explicitly think of using it as a standalone mode anymore as long as you’re using Computer), furthering the state of the art significantly
-
Auto-review becomes default for new users with 97% accuracy
By
–
Auto-review is now the default for all new users. A classifier subagent reviews actions in context before deciding whether to allow, block, or ask for approval. Our evals show it's 97% accurate, with most misses near ambiguous edges.
-
Flywheel of models and data limited to verifiable domains
By
–
Models, synthetic data, and environments are a flywheel now. The best model generates the best data, trains a better model, generates even better data. Rinse and repeat. But the flywheel only spins where you can verify: math, code, agents with graders. Everywhere else it merely