Joel-on-Software style rants and reflections on running a disciplined software practice in 2025.
Why OpenClaw matters: prompt-programmable workflows that run in your own channels with memory, automation, and no extra deployment stack.
A practical fix guide for the OpenClaw error: Agent failed before reply: OAuth token refresh failed for openai-codex.
A practical cautionary tale about accidentally enabling deep thinking mode in OpenClaw and draining token budget faster than expected.
Why building personal projects on your phone with OpenClaw beats doom scrolling and outperforms desktop-only productivity plans.
A practical deep-work protocol: remove phone proximity and use a 90-minute timer to reduce cognitive drag and increase output.
AI can now translate apps across frameworks and databases fast enough to change software strategy.
Most AI analytics is still polished reporting. The real shift is conversational reasoning over live operational data.
Single-agent coding feels fast until it confidently gets something wrong. Pairing agents catches errors early and improves output quality.
A practical look at which AI companies can survive ongoing losses, and why distribution and cash discipline matter more than model demos.
AI agents make software creation cheaper, but Jevons Paradox suggests demand rises even faster and shifts value to judgment.
Why social platforms should expose MCP servers so humans can safely delegate drafting, editing, and publishing to agents.
AI makes tiny personal software practical again: build one-off apps in a morning to solve your own real-world itch.
AI turns user stories into working drafts fast; the bottleneck shifts from specification throughput to taste and judgment.
Run a daily app-analysis cron that proposes one scoped feature, then reply yes/edit and auto-ship to a GitHub test deploy.
A lunchtime Shortcut + OpenClaw webhook loop that turns reminders into quick wins without building a full app.
Why live prototyping with stakeholders beats story-heavy "fake agile" workflows that behave like waterfall.
A practical pattern: daily prompt cron + quick reply + automated commit/push + CI/CD auto-deploy.
A quick note on how letting AI draft code while you review the output gives you the best of both worlds.
Quick view of today's date and how far through the month we are.
Looking for the live jobs prototype mentioned in essays?
Short, practical rants written between consulting calls—opinionated takes on AI tooling, software delivery, and developer ergonomics.