airwolf is that you?

can you share here so we can copy? jk jk i actually have something cooking (same 3 month dev cycle) and just submitted it to codex plugins and claude connecters yday. on the SEO AEO front, ive learned how to automate a lot of that by practicing on cheap content sites, and now have directed those skills to the app i've built. happy to trade notes (and share code repo) if you're interested!

We know the first wave of attacks were directed at IDF installations not at kibbutzim and nova. The second wave was directed was wanton violence. If you read Hebrew twitter after October 7 it was common to see people bring up how the attack benefited Netanyahu and cast suspicion on Shabak and their use of Mustaribeen to fan the flames, but not realizing how quickly it’d get out of control

depending on age of your kids, they might like printed and laminated checklists (created by GPT, of course). big font, big checkboxes, and one per part of the day, themed and customized based on their interests.

2026 and Israeli Rabbis still forbid Jews from ascending the Temple Mount. This ruling has not changed for decades.

dubai duty free has you covered my friend dubaidutyfree.com/millennium…

cue the SAFe Agile Release Train Engineer

can in introduce you to something i built that helps with this thrashing problem with long running tasks? im about to launch it today, its been work in progress for 2 months but Astra really helped push it across the finish line. screenshot below of a long running 20 hour task that was completed autonomously.

love the michael knightt vibes

The run took 1.02 billion tokens. 99.7% of the Claude-side tokens were cache reads because the process keeps every agent conversation append-only. On subscription that's $0; at API list prices it's about $235, or $17 per merged PR at the new cache-read rate

my team of agents completed a 21-hour agent loop comprising of thirteen core outcome driven tasks tickets follow a very specific determinstic flow, were co-created by me + agent, authorized after some back and forth, and then autonomously executed. fable orchestrated codex sub-agents as workers, absorbing two driver outages (a model-capacity failure, then a weekly-quota exhaustion) without stopping for 21 hours, and after human verification, fourteen PRs merged to main

that feeling of running out of all claude and codex credits because i created a system to for my agents to 20+ hours straight to create healthy, functional PRs