sqlite-utils 4.0rc2, mostly written by Claude Fable (for about $149.25)

2026-07-05 · Source: Simon Willison's Weblog · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Advanced, long

Summary

sqlite-utils 4.0rc2 has been released, with significant development assistance from Claude Fable, costing an estimated \$149.25. This release candidate addresses critical issues, including a severe `delete_where()` bug that could lead to data loss by failing to commit transactions. The new version refines its transaction model, ensuring all write methods like `insert()` and `upsert()` now commit automatically unless explicitly managed by `db.atomic()` or `db.begin()`. Key changes also include `db.query()` executing SQL immediately and committing `INSERT ... RETURNING` statements without requiring iteration. Compatibility with Python 3.12+ `sqlite3.connect(..., autocommit=True/False)` options was also ensured. The development process involved 37 prompts, 34 commits, and over 1,500 code changes across 30 files, with GPT-5.5 xhigh also contributing to a final review.

Key takeaway

For software engineers developing or maintaining database utilities, this case demonstrates that integrating advanced AI models like Claude Fable and GPT-5.5 into your development workflow can dramatically enhance code quality and accelerate release cycles. You should consider using AI for critical pre-release code reviews and even for generating routine documentation. This approach helps catch subtle, high-impact bugs, like transaction handling errors, before they reach production, improving reliability and reducing manual effort.

Key insights

AI models can effectively identify and resolve complex, critical software bugs and design flaws.

Principles

• AI models can serve as effective code reviewers.
• Multi-model review enhances bug detection.
• Automate release note generation.

Method

The author used Claude Fable for initial review, iterative prompting (37 prompts, 34 commits), and then GPT-5.5 xhigh for a final cross-model review to identify remaining issues.

In practice

• Use AI for pre-release code review.
• Employ multiple AI models for validation.
• Outsource routine documentation like release notes to AI.

Topics

• sqlite-utils
• Claude Fable
• GPT-5.5
• AI-assisted Development
• Transaction Management
• Software Release Engineering
• Code Review Automation

Code references

• simonw/sqlite-utils

Best for: Machine Learning Engineer, Entrepreneur, AI Engineer, Software Engineer, Director of AI/ML

Related on AIssential

• GPT 5.4 is a big step for Codex — GPT 5.4 marks a practical leap for AI agents, excelling in reliability, speed, cost, and precise instruction following.
• Claude Fable 5 is here! — Claude Fable 5 offers unparalleled performance for complex, long-horizon tasks at a premium cost.
• Production-Grade Text-to-SQL Agent with Claude Code, LangGraph, Langfuse, FastAPI and Qdrant — Building robust Text-to-SQL agents requires RAG, explicit state machines, and rigorous evaluation to prevent hallucinations and ensure reliability.
• 🎙️ How I AI: Claude Fable 5 review & How Braintrust uses AI agents, evals, and CI to ship better software — Optimal AI deployment requires matching model intelligence to task complexity, supported by rigorous evaluation and feedback loops.
• Claude Fable 5 + GPT-5.5 = GOD MODE — Combine Claude Fable 5's superior planning with GPT 5.5's cost-effective execution to mitigate high costs and rate limits.
• Claude Fable 5 TOMORROW? GPT 5.6 Kindle, OpenAI IPO News, Gemini 3.5 Pro, Nex-N2, & More! AI NEWS! — The AI landscape is rapidly evolving with major model releases, IPO filings, and agentic advancements across key players.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.