I orchestrate AI agents. Building Session Orchestrator to plan parallel sessions, check the results and pick up where the last session ended.
Wien, Österreich
- Tweets 486
- Following 295
- Followers 127
- Likes 994
Pinned Tweet
I orchestrate AI agents across tools and projects.
I'm building Session Orchestrator around that work: plan the tasks, run parallel sessions, check the results.
Each session needs a useful handover: what changed, what was checked, what's still open.
6
436
Bernhard Götzendorfer retweeted
Peter Steinberger, the creator of OpenClaw, open-sourced his entire agent setup
The core idea is one folder that every agent on his machines reads from.
A single AGENTS.MD holds his rules, and a script symlinks it into ~/.claude/CLAUDE.md and ~/.codex/AGENTS.md
This way Claude Code and Codex always follow the same instructions.
Every other repo gets one line at the top:
READ ~/Projects/agent-scripts/AGENTS.MD BEFORE ANYTHING.
Change a rule once and every project picks it up.
Skills work the same way. 69 of them live in one place, each with a short description the agent reads to decide what to load, and scripts/sync-skills links them into both agents.
Fork it, replace his rules with yours, and delete the skills you do not need. His AGENTS.MD is full of his own hosts and accounts.
6.6k stars, MIT - github.com/steipete/agent-sc…
78
193
1,614
257,892
Bernhard Götzendorfer retweeted
Claude Sonnet 5.5 just ONE SHOT this zombies game.
Max effort is noticeably better than every other level but it is expensive.
This game cost $177 and took 49 minutes to build.
This is insane.
65
39
1,099
60,053
I took first place in Track A1 at HACK_002 this weekend with Pflegegeld-Prüfer, an Austrian care-allowance checker.
For this one, I tried a new setup: one lead agent coordinating several coding sessions. I mostly talked to that lead, tested the app, fed back what wasn't working and tried the next version.
The idea was already worked out before I arrived. On site, the work was turning it into something I could actually show, while figuring out the presentation and recording this walkthrough.
The video is the actual submission. Me explaining the app, with the agents handling the edit. It includes the results that still fell short, too.
I'm still building on it.
4
2
137
dang
Antigravity 2.0 now features a dedicated planning mode, just like the Antigravity CLI.
Type /plan and the agent steps back to think through the task, and conduct extensive research, before generating an implementation plan for your review. The agent will ask for your approval before it begins executing on the task.
You can also ask for a plan naturally in your prompt, for a lighter version of a plan.
2
2
119
Real question for everyone on a 16gb gpu:
What stops / annoys you the most from using local LLMs every day? Tiny context? Stupid small models? Low tok/s?
I'm planning the next crow-nest build for exactly your setup. tell me what you want / need. Would appreciate a like / repost so more ppl with 16gb gpu see this.
#RTX #AI #LLM #16GB
3
2
7
203
Bernhard Götzendorfer retweeted
OpenClaw deleted around 400k LOC of its own tests without much change in code coverage. Modern models just love writing tests for every tiny change, even if they aren't useful. This skill helped. github.com/openclaw/openclaw…
230
266
6,707
1,272,046
Bernhard Götzendorfer retweeted
The agent loop in a coding agent is tiny.
Model call → Tool call → Append result → Repeat.
Everything around that loop is the harness that turns it into a useful coding agent.
11
9
32
2,863
Some of the work in a Session Orchestrator brief was already done.
Checking the current code before starting made that visible. The brief alone wasn't enough to decide what to build next.
1
36
AgenticCutter has been through two App Store rejections.
The direct release and the Store submission are separate jobs. I can get one out and still have work left on the other.
34
