Labs

Thrivbe is an open lab. Alongside client work we run several experiments a day — with frontier AI models, autonomous agents, and new ways for people and machines to collaborate. This notebook documents them as they happen: the hypothesis, what we built, what we learned. No hype — just the work, in the open.

In progressAugust 2026

Can a sealed agent fleet trade unsupervised — on paper?

Paper-Trading Agent Fleet

Self-editing loop patterned on autoresearch by Andrej Karpathy

Labs

Can a sealed agent fleet trade unsupervised — on paper?

Can you see your whole network as one living graph?

Can a network tell you which two organisations should meet — and can you prove it's not guessing?

Does a knowledge graph remember a community better than flat search does?

Can a website's voice agent remember returning visitors — and should it?

Can a model read an entire physical book from a shaky phone video of someone flipping through it?

Can an agent direct a finished film if the human only ever answers one question at a time?

Can a spreadsheet join your meeting and answer questions out loud?

Can an agent build 25 genuinely different websites without a human reviewing any of them?

Does a pre-LLM agent architecture still have anything to teach a kernel built on LLMs?

Can an open-source prompting skill direct a multi-shot AI film with zero re-rolls?

Which of your automated jobs are quietly broken before you leave for a week?

Can a phone-only operator command the coding agents on their laptop?

Can an openly-AI employee character build a real B2B funnel on short-form video?

Can a voice agent host an entire website — answering questions, navigating you to any page, and even driving an interactive network graph?

Can you spin up a whole AI company for one person?

Can a factory of AI agents launch products — and kill them — on real revenue signals?

How far can you get on the free tier of every LLM?

Can an AI be its own data-protection officer?

Can an agent organisation improve itself while you sleep?

Can a festival in a field run software with no signal?

What does a knowledge base look like when its main reader is an AI?

Do agents need their own computers?

Can you run your company by talking to it?

What does a task manager look like when agents are first-class users?

Can an AI-generated world engine run entirely in the browser?

Which LLM router actually picks the right model?

Can we convert static inspirations into living systems?

Can an agent watch you work and write your content for you?

Can your laptop defend itself from its own agents?

What did we find when we security-audited an AI sales agent?

What does the cockpit of an AI company look like?

Can agents run the whole sales motion around one human?

Can one person run a whole company as an operating system?

Can an agent crew take a video from idea to publish?

Can agents find each other, prove identity, and request services over the open web?

Can agents keep watch on the creator ecosystem for you?

What changes when your AI remembers every conversation?

What does an AI organisation's org-chart look like?

Can AI manage a community's shared photo archive?

What fills your week — and what could AI take off your plate?

Can software make a volunteer camp actually low-effort?

Can an agent keep a backlog honest?

Can every meeting turn itself into decisions and tasks?

How small can a serious semantic search engine be?

Can you teach an AI taste?

What do ten years of connections look like as data?

Would you install an unaudited AI skill?

Can a voice twin answer for you?

Where do words actually come from?

What actually makes a community strong?

Can your website answer questions about you?

What does AI actually do to 143 million jobs?

Can AI rebuild a website pixel-perfect — and then take it over?

Join our Newsletter