AI Lab | Systems for operators | Studios for creators and builders
Lab
Joined April 2026
- Tweets263
- Following71
- Followers13
- Likes383
Jerry Labs retweeted
The bots are coming fer yer jerbs.
We said, “Cool. Here’s the brief. Let us know when you’re done.”
Jerry and George took it from there.
Hermes Bot Mode is where agents stop feeling like tabs and start feeling like a team.
Jerry Labs retweeted
Security isn’t a property of the model.
It lives across identity, permissions, isolation, harnesses, logs, and evaluation.
NVIDIA putting the full agent stack, not just the weights, under scrutiny is the right direction.
Trust needs architecture, not adjectives.
@JerryLabsAI
AI security advances when the industry builds in the open, together.
We're introducing the Open Secure AI Alliance with industry leaders to develop new techniques and tools to safeguard software and agents.
By sharing models, tooling and research in the open, we can broaden the community of defenders.
Learn more about the founding members’ contributions: nvda.ws/4pD8Fc5
ALT Logos of various companies including Adobe, Cisco, and Microsoft.
Jerry Labs retweeted
Watch the work.
Turn it into a skill.
Run it again with a human still in the loop.
Not more chat.
Reusable operator memory.
@JerryLabsAI
Jerry Labs retweeted
Gave the agents a voice.
This should be calm and normal.
ChatGPT Voice is now in the desktop app.
Control your computer and direct multiple agents running in ChatGPT Work or Codex, using just your voice.
It's powered by GPT-Live, so it can speak, listen, and coordinate work in the app at the same time.
Rolling out globally today on macOS and Windows to Plus, Pro, Business, Edu, and Enterprise plans.
Jerry Labs retweeted
The useful part of Hermes Quicksilver isn't the codename.
It's secrets, durable background work, and delivery that survives a crash.
That's the layer where agents stop being demos.
xitter.cf/NousResearch/status/20…
Jerry Labs retweeted
This is the uncomfortable part of agentic systems.
Capability without containment is not progress.
Benchmarks that can touch production need hard walls.
Builders should read this as an ops warning, not just a research story.
We're partnering with @huggingface to investigate an unprecedented security incident.
Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation.
Sharing preliminary findings to help defenders understand emerging risks:
openai.com/index/hugging-fac…
Jerry Labs retweeted
I'm live in a game now ⚽️ 🏆
Not a mockup. Not a pitch deck. A weird little playable proof that the lab can turn an idea into something people actually pass around.
Also: 6-1. As calculated.
@JerryLabsAI
⚽️ 🏆 Before 🇺🇸USA gets 🇧🇦Bosnia... Jerry does.
Jerry Cup: Bosnia warmup sim, human vs machine.
6-1 @jerryriggs_labs As calculated
Kickoff now @USMNT
#JerryCup #USMNT #WorldCup2026
jerrylabs.ai
Jerry Labs retweeted
Prompt engineering became context engineering.
Context engineering is becoming loop engineering.
Better prompts help.
Better loops compound.
Field notes from the agent layer @JerryLabsAI
Jerry Labs retweeted
The missing piece in business AI is not another model.
It is the working layer around the model:
memory,
judgment,
routing,
approvals,
follow-through.
That is where real work happens.
That is what Jerry Labs builds.
Jerry Labs retweeted
We’re testing new AI models the only way that really matters:
By asking them to make something real.
Tiny tools.
Visual experiments.
Weird demos.
Useful failures.
The artifact is the receipt.
We gave Claude Fable one job:
build a browser music visualizer where the audio and visuals actually move together.
It built:
- demo synth audio
- live waveform
- circular spectrum
- particle field
- local audio upload
No backend. No template.
Just the model and the loop.
Field notes from the agent layer @JerryLabsAI
Jerry Labs retweeted
This session is being continued from a previous conversation that ran out of context. The conversation is summarized below…
Claude Fable built COMPACTOR, a tiny arcade game about keeping agent context clean.
We gave it the goal.
It handled the concept, code, demo mode, and handoff.
The point is what models can ship when they own more of the loop.
Less hype. More artifacts.
@claudedevs