CompleteTech

[CompleteTech @CompleteTechLLC]

@thsottiaux I write meta prompts with goals and subgoals, seems to handle large scale projects pretty easily. Working on this one for nearly 2 weeks now.

[CompleteTech @CompleteTechLLC]

@thsottiaux Wish I saw this sooner, /goal best thing since the Claude code source code accidental leak tbh.

Always, it is the way.

[CompleteTech @CompleteTechLLC]

@BunsDev Here kitty kitty, here kitty kitty.

[Y Combinator @ycombinator]

We're excited to announce Peter Steinberger as a speaker at Startup School 2026!

@steipete is the creator of OpenClaw, the open-source AI agent that went from a weekend project to the most-starred software repo on GitHub in under 5 months, with 346k+ stars. He's now at OpenAI building the next generation of personal AI agents.

https://t.co/xVr1OXFRLY

[CompleteTech @CompleteTechLLC]

@vincent_koc @openclaw @SFMTA_Muni ABC - Always Be Coding

[CompleteTech @CompleteTechLLC]

@steipete Your talk on this was so validating to a thought process I suggest frequently. "Spend more time building tools than agents."

[OpenClaw🦞 @openclaw]

Come join our watch party!! https://t.co/ZzceY1emHk

[CompleteTech @CompleteTechLLC]

@vincent_koc Vincent, you're so fast on this! 1 minute and repost. We just need to follow you for the latest on OpenClaw imo.

[CompleteTech @CompleteTechLLC]

New day, New Model.

[Boris Cherny @bcherny]

Claude Opus 4.8 is out today. It's our strongest coding model yet: up on SWE-bench Pro (from 64.3 to 69.2) and noticeably more honest about its own work. It tells you when it's unsure and catches its own bugs instead of declaring victory early. Same price as 4.7.

[CompleteTech @CompleteTechLLC]

@mitsuhiko Great list of changes, can't wait to give it a shot.

[CompleteTech @CompleteTechLLC]

@MatthewBerman This could be a problem. lol - Bots will start doing this now.

[CompleteTech @CompleteTechLLC]

@vincent_koc I got 99 problems but unsupported agents ain’t one.

[CompleteTech @CompleteTechLLC]

This Memorial Day, I’m honoring the brave few who endured 12 days, 14 hours, and 36 minutes of /goal chaos, #codex resets, and unwavering persistence.

@thsottiaux and @sama — your service will not be forgotten.

May the goal persist until completion.

[CompleteTech @CompleteTechLLC]

@steipete @Cloudflare The actual setup was per-project code-server browser IDEs backed by persistent containers, shared ingress/TLS, mounted repos, Codex/gh/SSH state, and browser-accessible terminals. Not Firecracker, but same intent from a local/private-cloud direction.

[CompleteTech @CompleteTechLLC]

@steipete @Cloudflare Interesting! I was trying to solve the same workspace + compute problem from the local side and ended up at a similar shape.

[CompleteTech @CompleteTechLLC]

@cybersecdev_ @thsottiaux You can solve this with a meta prompt. Feel free to reach out if you have any questions.

[CompleteTech @CompleteTechLLC]

Very helpful information if you have a long running /goal.

Token usage: total=48,236,955 input=42,226,324 (+ 3,381,871,360 cached) output=6,010,631 (reasoning 1,513,033)

[ClawDaddy @clawdaddysh]

Heads up for those of you running /goal::

If Codex .133 breaks /goal with thread/goal/get failed, ask your agent:

“Check codex app-server daemon version; restart stale Codex TUI/app-server processes if appServer != CLI; verify ~/.codex/goals_1.sqlite; restore the goal row from session JSONL only if missing.”

[CompleteTech @CompleteTechLLC]

@mofrymatic "Beat the game no heuristics, no cheating"

[CompleteTech @CompleteTechLLC]

What is the longest running prompt you have had?

[CompleteTech @CompleteTechLLC]

@doodlestein I'm catching up.

[CompleteTech @CompleteTechLLC]

How many tokens did you use?