sqlite-utils 4.0rc2: Written by Claude Fable, $149.25, Fixes Data Loss Bug

Simon Willison shipped sqlite-utils 4.0rc2, a Python library for SQLite, written mostly by Claude Fable. The total cost: about $149.25 in API fees over 37 prompts and 34 commits, changing +1,321 -190 lines across 30 files.
What Changed
The release candidate fixes a critical data loss bug in Table.delete_where() — previously, it never committed, leaving the connection in-transaction and poisoning all subsequent writes. Here's the reproduction from Fable's analysis:
db = sqlite_utils.Database("dw.db")
db["t"].insert_all([{"id": i} for i in range(3)], pk="id")
db["t"].delete_where("id = ?", [0])
# conn.in_transaction is now True
db["t"].insert({"id": 50})
db["u"].insert({"a": 1})
db.close()
# Reopen: rows are [0, 1, 2] — the delete, row 50, AND table u are all gone.
Willison calls this a “really bad bug” — one he’s glad Fable caught before the stable 4.0 release.
New Transaction Model
Every write method (insert(), upsert(), update(), delete(), delete_where(), transform(), create_table(), enable_fts(), etc.) now runs inside its own transaction and commits before returning. No manual commit() needed.
Two exceptions:
- Use
db.atomic()to group multiple operations atomically - If you open a transaction with
db.begin(), the library will never commit it — you control the commit
Python 3.12+ autocompat Support
Fable also ensured compatibility with Python 3.12's new sqlite3.connect(..., autocommit=True|False) mode. The library's automatic per-method transactions are designed for default connection mode, and Fable patched the test suite to pass under the new autocommit settings.
Review Process
Willison started the work on his iPhone via Claude Code for web, then switched to a laptop for final review via GitHub PRs. He says reviewing documentation edits first is an excellent way to understand what changed. A final review by GPT-5.5 caught additional edge cases — the full PR transcript is shared.
Who It's For
Python developers who use sqlite-utils for lightweight database ops and want a production-ready 4.0 release with proper transaction safety.
📖 Read the full source: HN AI Agents
👀 See Also

OpenClaw's QMD Memory Search Fast Path Had Silent Bugs
OpenClaw's built-in memory search uses basic keyword matching, but users can switch to QMD for semantic search across workspace markdown files. A fast path through MCPorter was broken with three bugs causing every call to silently fail and fall back to slower CLI execution.

OpenClaw Skill Reduces Accessibility Tree Tokens from 600K to 1.3K
A developer built an OpenClaw skill that uses ML-based element ranking to prune accessibility trees, cutting slickdeals.com from ~598K tokens to ~1.3K tokens by keeping only the top ~50 actionable elements.

Open-source tool for AI-curated Reddit feeds using Cloudflare, Supabase, and Vercel
A developer open-sourced a self-hosted tool that filters Reddit for quality posts about AI-assisted development, using Cloudflare Workers for cron jobs and proxies, Supabase for storage, and Vercel for the frontend. The tool includes engagement scoring, optional LLM summaries, and costs $1-2/month for AI processing.

Double-Buffering Technique for LLM Context Windows Eliminates Stop-the-World Compaction
A technique called double-buffering can prevent LLM agents from freezing during context window compaction by summarizing early and maintaining two buffers, allowing seamless handoff at no extra inference cost.