Software & DevOps

simonwillison.net

Simon Willison upgrades datasette-checkbox after a permissions API change. He first reproduces a real page failure that the existing tests miss, adds a failing page test and gives Codex the runner plus upgrade guide. After the agent changes the permission call, Willison checks the interactive feature and adds separate assertions for root and non-root users.

Codex

simonwillison.net

Simon Willison combines two existing tools into a terminal-to-HTML sharing page with Claude Code for web. He references the repository's rich-text conversion and GitHub Gist authentication implementations, tries a pull-request preview, then checks the published tool in Firefox and Chrome. The recording includes failing tests and a preview-domain authentication limitation rather than a clean one-shot success.

Claudevia Claude Code

punkfab.com

Claude Code built circuit viewers and real-server tests that caught a host form bug; published plugins retain visible routing failures and explicit host limits.

Claudevia Claude Code; Codex supplied separate upstream patches

linkedin.com

Claude Code extended Ralph through its own queued plans, adding worktrees and review loops while Stevens retained architecture choices and blocked-task stops.

Claudevia Claude Code inside the Ralph task loop

Andrew Leech coordinated Claude sessions to rewrite an MCP plugin for MicroPython, profile idle TLS polling and implement a shared wake-source mechanism.

Claudevia Claude Code

Kian Kyars coordinated Codex, Gemini and Claude workers on a Rust SQLite-like engine, using task locks, shared notes and sqlite3 checks while noting limits.

Codexvia Codex CLI alongside Gemini CLI and Claude Code

letsgochristo.com

Christo Roberts used Codex to ship RepoFinder on Cloudflare Workers, verifying contracts in a temporary clone and adding recommendations with visible fallbacks.

Codex

devpost.com

Codex packaged LiDAR timing diagnostics with synthetic fixtures, deterministic rules and evidence exports; the original field investigation stayed human-led.

Codexvia Codex with GPT-5.6 Sol

closetheloop.dev

The ccc-morph author used Codex to build a PTY wrapper that remaps shortcuts per program, preventing a double Ctrl-D habit from closing Codex and the shell.

Codexvia Codex CLI

lakshyaag.com

Agarwal reports using Cursor and two models to build a working Go terminal agent with file tools and shell commands despite having no previous Go experience.

Cursor

linkedin.com

Cursor completed recruiter-agent integrations after a broad CrewAI-to-LangGraph rewrite failed; Karpić tested it and supplied a working foundation by hand.

Cursor

linkedin.com

Cursor reproduced authenticated REST failures using Swagger readiness, JWT tooling and curl; Stevens refined its stored workflow until the fourth bug ran alone.

Cursor

github.com

Devin added opt-in Statsig replay with client text and input masking, startup-failure isolation and tests; the merged PR retained an unchecked lint item.

Devin

github.com

Devin found production modules hidden from a dead-code analyzer by obsolete tests, checked scripts and workflows, then removed the orphans in a merged PR.

Devinvia Devin cloud workspace

github.com

Devin batched CRM duplicate candidates, preserved pair cursors and archived-chain behavior, and merged a patch with a scoped production query measurement.

Devin

github.com

Devin built a repeatable MCP latency benchmark with direct-call baselines, percentile reports and a scheduled workflow using synthetic secrets for local runs.

Devinvia Devin cloud workspace

github.com

Devin traced a mailbox test flake to a stale continuation alarm, deleted it only when no work remained and added tests preserving overdue repair alarms.

Devinvia Devin cloud workspace

note.com

Infomart used Devin's Python analysis to map API and SQL load, then hit private-dependency limits during generated test execution and moved reports to Slack.

Devinvia Devin cloud workspace

github.com

Devin overlapped MCP admission with a pure registry read, kept execution and writes behind approval and added denial coverage before the Kody PR merged.

Devinvia Devin cloud workspace

github.com

Devin traced Chrome font flashes to CSS font-face rebuilding, pinned loaded JavaScript twins and added lazy-loading, retry and browser-edge-case coverage.

Devinvia Devin cloud workspace

devin.ai

Devin scanned Dioxus dependencies and Cognition’s websites, exposing unnecessary build features and metadata defects with qualified local and shipped outcomes.

Devin

github.com

Devin traced an Electron Windows release failure to PowerShell and npm argument forwarding, merged a script repair, and kept the later rerun claim separate.

Devin

github.com

Devin fixed a ready-then-deleted Vectorize race with bounded metadata retries and a regression that failed against the old implementation before PR merge.

Devinvia Devin cloud workspace

github.com

Devin added branch provenance to preview deployments and guarded cleanup with PR-state rechecks, keeping ambiguous or newly deployed resources before merge.

Devinvia Devin cloud workspace