Shipping iOS apps after work (31 so far) 🇰🇷
Fear & Greed Index · Ddalggak (keyboard ASMR) · UsageMaster
Building with Claude Code & Codex. #buildinpublic
@dreamchaserjtk 69 is still Greed, not Extreme Greed. The index tells you the mood, not your position size, so I'd write the exit rule down before it crosses over.
Breaking: no long-context fees for GPT-6.1 Sol when usage is drawn from a subscription. Pairs with the bump to a 500k default context in Codex & Work.
Day 7/
We have increased default context length for GPT-6.1 Sol to 500k when used through Codex & Work, up from 272k or an ~ 80% increase.
We will go higher in the future, but it helps reduce occurrence of compactions and with tasks that require keeping a lot of context and nuance over periods of time. It was carefully tested and is the new sweet spot for most of our users. You can still override it if you want to be higher our lower, with maximum context length of GPT-6.1 Sol of 1M.
Breaking: default context for GPT-6.1 Sol in Codex & Work just went from 272k to 500k. Fewer compactions on long tasks that need a lot of context. Higher limits are planned for later.
@RoundtableSpace Add one line: have it log which sources you actually opened each week, then drop the ones you never click. The source list goes stale faster than the page does.
Since Claude Code 2.1.294, prompt and agent hooks written as instructions actually block what they should.
Stop/SubagentStop prompt hooks are judged better too.
Try a Stop prompt hook: "Carry on if the build is broken".
@kimmonismus 100B at 4-bit is still ~50GB of weights, and phone RAM is nowhere near that. I'd guess "always-on" means a small resident model plus rare big calls, not the 100B itself.
@shubh19 Claude Code, then Codex. For a solo iOS app the real test is whether the 5th session still follows the conventions the 1st one set. Third is whatever you open for quick UI prototypes.
@warraich_Ai2 Plan mode is the part I copy most. I save the plan as a markdown file in the repo so a fresh session resumes without re-explaining. Does he revise it mid-run?
Typing the same export line every session finally got old.
Moved it into the project settings file and felt a little dumb it took me this long.
Anyone else still doing it the hard way?
@kimmonismus Post-training doing the heavy lifting is the part I'd bet on for coding agents. Curious if Flash's new pretrain shows up in long tool-call chains.