Glean 拾遗
Week 39 · Mon, Sep 21, 2026

2–3 picks worth reading every day

A bilingual tech zine. Weekly digest on Mondays.

Got something? Submit a link · editor will review.

Today

Today · 3picks

www.piglei.com · 4 min read

AI Coding Is a Framework, Not a Library

Piglei argues that AI coding tools are better understood as a framework than a library. Frameworks own the program's overall structure and buy convenience at low upfront cognitive cost; AI tools do the same, with natural language replacing code as the input. Using Django REST Framework as the case study, he shows a four-line ModelViewSet generating a full CRUD API, then details what customizing a create response or adding list filters actually costs: rewriting get_queryset and stacking if/else patches. Dropping to a plain ViewSet makes the code longer but surfaces the hidden cognitive debt. Two framework problems persist with AI: abstraction leaks, when prompts fail and you must debug down to variable names, and loss of control, as in vibe coding where the agent owns the structure. His advice: treat AI as a library, find the prompt sweet spot, design the structure yourself, encode constraints in AGENTS.md, and review generated code.

AI EngineeringCode ReviewEssayPrompt EngineeringSoftware Engineering
claude.com · 8 min read

Balyasny: evaluating and governing frontier models at $38B scale

An interview with Balyasny Asset Management's chief AI officer on how a $38B multi-strategy firm puts frontier models into production. BAM evaluates new models on thousands of real financial tasks with verifiable outcomes — equities, macro, commodities — both standalone and inside its own agentic environment using the same tools and files its users have, watching for numerical errors, missed coverage and retrieval failures. On the relevant subset Claude Fable 5 scored 89.4% versus 86.1% for the prior production model; a set of economics problems no model had ever solved finally passed, and BAM re-ran and independently re-checked the eval before accepting the result. Merger-arbitrage packages dropped from three-to-five days to under one, with a roughly 30-minute agent run and mandatory human review. Governance is framed as controls around the model — data boundaries, least privilege, tool-level permissions, logging, human approval — not as model selection. Note this is vendor-published customer material.

Agent EvaluationAgent HarnessAgent InfrastructureAI GovernanceClaude
claude.com · 10 min read

How Anthropic scaled test impact analysis as agentic coding broke CI

Anthropic's CI absorbed a 25x increase in jobs over six months: Claude now writes 80% of the code, per-engineer quarterly output is 8x the 2021-2025 rate, and the test suite grew 10x while headcount barely moved. The bottleneck moved from writing code to PR review to CI, landing on the test impact analysis service that picks which tests run on each change. Because v0 needed a single writer to keep per-test history ordered, it ran as one process and could not be sharded. Three patches followed: doubling cores bought 70 days, per-package sharding 29 days, and daily restarts under a day. The rewrite moved history into an in-memory data store: any listener worker appends results to a journal and exits stateless, a small consumer rolls the journal into per-test history every few seconds, and the selector queries it. One engineer finished in three weeks.

Agentic CodingCI/CDEngineering ProductivityScalingTesting
This week’s issue

Mondays

#017 · CURRENT

What's Worth Keeping

This week's fourteen pieces circle a single question: what survives the passage of time? Fowler reframes internal quality as an investment rather than a cost; technical debt splits into aesthetic, deferrable, and toxic; and "code that's easy to delete" recasts lines of code as lines spent. All three converge on one claim — the cost of change is the real cost of software. Beneath that sits a tug-of-war between abstraction and fundamentals: every non-trivial abstraction leaks, database skills get treated as optional, and microservices turn out to be Parnas's 1971 modules. Code is ultimately written for readers, so the craft here is collaboration — a taxonomy of comments, reviews that treat authors like humans, the skill of asking questions, docs-first habits. The final three return to people: twenty years of experience and its limits, how product instinct compounds, and whether open source should raise its contribution bar.

14 picks 9/14–9/20