Code, Think.
Putting AI coding into real work: Claude Code, Codex, agents, code review and automation. Written by someone who ships.
LATEST ENTRIES
ENTRIES: 06CyberToken-S: Squeezing More Out of Free Credits
Pool the tokens platforms give away, plug them into OMP + Magpie, and put together a cyber crew of your own. Codex leads; free models do the subtask work.
I Made Claude and Codex Hand Off Tasks to Each Other
Claude's 5-hour window runs out, but Codex still has quota. To keep long tasks running without interruption, I built a session handoff workflow so two AI coding agents can read each other's records and pick up the same task.
I Put Two AIs in My Pre-commit Hook and Made Them Argue
Three layers in my pre-commit — image compression, formatting detection, and a Codex + Claude adversarial review. Codex attacks first, Claude defends. If both agree something's wrong, the commit is blocked.
I Bolted an Argue Bot onto Our Feishu Code Review
This is not about fighting review with AI. Our team's review system is genuinely well-designed — I just wanted one more layer after the Feishu report lands, so I can see what's worth arguing before I even decide to context-switch.
Why I Push the First Claude Code / Codex Window Into the 7am Slot
This is not really about the automation script. It is about budget reallocation: for someone like me who often gets to the office at 10:30 or even 11:00, pushing the first window into the 7am slot turns the workday from roughly 2 reachable windows into 3.
Built an AutoGLM Bot to Clock In for Me. Ran It for Three Months.
The motivation was simple: nobody at the company is obsessing over attendance, but late arrivals and missed punches still sit there in Feishu and look bad over time. So I used a spare Android phone, ADB, and AutoGLM to automate the whole thing.