! TOP NEWS !
OpenAI's CRO sent an internal memo trashing Anthropic. Called them a "single-product company" and accused them of inflating their run rate by $8B 📄
The Fed summoned Wall Street CEOs over an AI model's cyber capabilities. First to clear a full 32-step corporate cyber range 🏦
Andon Labs gave an AI $100K and a 3-year lease in SF. It hired employees, sourced contractors, and opened a boutique. The mistakes are peak AI 🏪
! PRODUCT NEWS !
Anthropic shipped three things at once. Claude Code desktop app, "Routines" that finally runs while your laptop sleeps, and Opus 4.6 hit 97% of human-level alignment research 🧠
MiniMax dropped a 230B open-source model that wrote its own training code. Already the top performer for agentic local tasks 🤖
GPT-5.4 Pro just solved a 60-year-old unsolved math conjecture. Built a rigorous formal proof in under 80 minutes 🧮
! FOR THE CURIOUS !
🚀 Berkeley scored 100% on every major AI benchmark. Without solving a single task.
UC Berkeley's RDI lab built an exploit agent that hit ~100% on eight of the industry's most cited benchmarks. SWE-bench, WebArena, GAIA, all of them. One exploit was a 10-line Python file that forced every test to pass.
Models have already been caught copying answers from git history. OpenAI retired its own benchmark after finding 59.4% of tests were broken. If you're picking models by leaderboard score, Berkeley's full breakdown is worth 15 minutes.
🕸️ Ever tried to make an AI agent scrape a site with a login wall?
AI agents like OpenClaw and Claude Code are powerful. But they can't use the web beyond basic search and fetch. The moment they hit a dynamic page or Cloudflare, they fall apart.
TinyFish built the fix. Full-stack web infra for AI agents, all under one API key.
87% fewer tokens than MCP per operation
#1 web agent on Mind2Web (89.9% accuracy)
Scales with parallel browser sessions at enterprise volume
Works with Claude Code, OpenClaw, Cursor, and other agents
And the fastest way in is their CLI + Skill: install the CLI (npm install -g @tiny-fish/cli), drop a single markdown Skill file into your agent's directory, that’s it!
! BANGER EVENTS !
Ship to Prod Hackathon - SF
Build real AI agents with AWS, Wundergraph & more.
🏆 $45k+ in Prizes!
🗓️ Fri, 24 April . 9:30 AM PT
AI Agents in Prod - Ship Demo Day - SF
Showcase your project or attend workshops. Backed by Google Deepmind, Eleven Labs & more.
🗓️ Thurs, 23 April . 9:30 AM PT
Upscale Conference - SF
Top creators share what actually works through panels, hands-on workshops, and real connections.
🗓️ June 3 - 4 . 9:30 AM PT
DigitalOcean Deploy Conference - SF
See how real AI companies run inference at scale. Leaders break down how they run AI in production
🗓️ Wed, 28 April. 12PM – 8PM PT
Interested in sponsoring an upcoming event or hosting your own?
reach out here
! OPPORTUNITIES WITH THE TEAMS WE LOVE !
Bland AI — Forward Deployed Engineer (SF)
Build and deploy AI voice agents directly with enterprise customers, from first call to production.
→ ApplyVercel — Software Engineer, AI SDK (NYC / SF )
Build and maintain the TypeScript toolkit behind AI SDK, powering AI-native apps for millions of devs.
→ ApplyDigitalOcean — Senior Forward Deployed Engineer, AI/ML (SF)
Embed with strategic customers to ship AI apps on GPU infrastructure and build tooling that scales across the platform.
