Daily Brief — 2026-09-22

Sources: AI Daily Briefing.md (X) + GitHub trending daily

Window: last ~24h (from 2026-09-21 07:42 BRT). Original posts only (exclude retweets/replies).

X credits before run: ~$8.49; after: ~$8.31.

X — signal

@poteto — Lauren Tan

• How she shipped 2,500 PRs to production last month (talk that was meant for Cursor Compile London; free on X). post · 6,822❤ · 504 RT
• Happy Grok 4.7 day (quote of the launch). post · 2,327❤ · 93 RT
• "i might be addicted to making games now." post · 747❤ · 11 RT

@rauchg — Guillermo Rauch

• Grok 4.7 on Vercel: 40% off; fast, smart, cheap — try on Gateway. post · 619❤ · 19 RT
• "Game recognizing game" (quote of Nat Friedman on Muse/OpenClaw). post · 498❤ · 6 RT
• Jev now callable over HTTP in AI Gateway (plus type-safe TypeSafe AI SDK API). post · 117❤ · 3 RT

@hwchase17 — Harrison Chase

• Jev as "decision models"; SemIf (open-source, Qwen3.5-based) free via LangSmith Gateway for a week. post · 119❤ · 10 RT
• Excited about decision models: agentic systems are good engineering around models ("build prod, not god"); Jev excitement shows people agree. post · 70❤ · 5 RT
• SemIf close to Jev on JevBench (75.4 vs 74.7); other OSS options exist. post · 76❤ · 8 RT
• Try Jev in LangSmith to score traces cheaply and accurately. post · 27❤ · 6 RT
• On AI checkpoints / multi-goal decision models: "kind of like goal but with multiple in a row." post · 71❤ · 11 RT

@simonw — Simon Willison

• Notes on Jev and system-one / decision models. post · 824❤ · 60 RT

@HamelHusain — Hamel Husain

• Pushback on "not a classifier" rage-bait (confidence scores / two-option choice still fit classifier). post · 100❤ · 1 RT

Silent (no original posts in window)

@martinfowler · @cursor_ai · @karpathy · @swyx

Gaps

@xAI — API not authorized to read this account's posts (skipped).

GitHub — top 5 daily trending

1. anthropics/financial-services (Python) — +424★ today — Reference agents, skills, and data connectors for financial-services workflows (investment banking, equity research, PE, wealth management).
2. agent-substrate/substrate (Go) — +498★ today — Agent Substrate: the core system.
3. dream-num/univer (TypeScript) — +202★ today — The Office Harness for AI Agents — Spreadsheets, Docs, Slides, Canvas, Relational Tables, and PDF in one runtime.
4. davila7/claude-code-templates (Python) — +33★ today — CLI tool for configuring and monitoring Claude Code.
5. google/ax (Go) — +2,324★ today — Google's open agentic orchestration runtime.