Hugging Face Becomes Platform for AI Agents

#1 𝕏

clem 🤗 says Hugging Face has become the go-to platform for AI agents, giving them access to 1 M HF Spaces to build and run the latest specialized models.

#2 📝 Simon Willison

Changes in the system prompt between Claude Opus 4.6 and 4.7 - A detailed look at how Anthropic's Claude system prompt changed between Opus 4.6 and 4.7, using their published system prompts as the basis for analysis. The post highlights the value of Anthropic publishing system prompts and links to deeper notes and artifacts used in the research.

#3 📝 Simon Willison

Claude system prompts as a git timeline - A research project that transforms Anthropic's published Claude system prompts into a git-based timeline, splitting monolithic markdown into per-model files with timestamped commits to enable diffs and blame. The artifact helps researchers trace prompt evolution and was used to author detailed notes comparing Opus 4.6 and 4.7.

#4 𝕏

xAI launched Grok’s Speech-to-Text API, delivering instant multi-speaker transcription across 25 languages at the market’s best price.

#5 𝕏

bolt.new launched the Bolt GitHub Connector (MCP), letting you copy-paste components from any open-source repo and even port entire codebases across languages—see a Swift project transform into TypeScript at 5:36.

#6 𝕏

Santiago reveals that Claude Code’s Esc+Esc shortcut rewinds your session to any previous checkpoint, letting you freely experiment with complex prompts. Now you can evaluate results risk-free and instantly undo unwanted changes.

#7 in

Peter Yang walks through a 16-minute live demo of Claude Design, showing how to build videos, slides, websites, apps and an initial design system.

Also covered by: @Peter Yang, @Greg Isenberg

#8 in

Greg Isenberg predicts a $1T+ market for agent-first startups that treat Salesforce, HubSpot and other SaaS as dumb backends, replace professional services with autonomous agents, and shift to outcome-based pricing.

#9 in

Anu Jagga Narang observes that instead of truly collaborative “multiplayer” AI efforts, teams are running isolated, single-player experiments on the same problem—weaponizing AI in parallel and never building the promised shared context or converging on a common artifact.

#10 𝕏

DeepLearning.AI highlights that AI tools like Be My Eyes boost independence for low-vision users by describing appearance and surroundings, but warns their subjective beauty judgments can spark confusion, insecurity, and psychological risks.

#11 in

Udi Menkes compiled 108 GenAI PM briefs into a 601-page wiki, uncovering Claude Code as the dominant PM tool in 2026 (85/108 mentions) and Guillermo Rauch as the most-cited influencer (53 mentions).

#12 𝕏

Teresa Torres highlights how Todoist’s AI team had to fine-tune their LLM prompt “temperature” so it wouldn’t drift into full marathon planning (when set too high) or lose its smart task‐capture flair (when set too low)—a balance they cite as one of their most pivotal AI deci...

#13 𝕏

Lenny Rachitsky relays @rabois’s insight that in the best organizations the CMO consumes the most tokens, highlighting marketing’s pivotal role in driving product engagement.

#14 𝕏

claire vo 🖤 says Claude Design shines as a production tool when you already have a polished design system from talented designers—but without one, it offers little value.

Also covered by: @Peter Yang, @Greg Isenberg

#15 𝕏

Guillermo Rauch predicts design will shift from manual tools to autonomous Design.md scripts run by coding agents, with teams spinning up personalized design capabilities via v0’s Platform API/Sandbox.