We poke at large language models, build GenAI apps, and write about what we find. No funding pitches. No hard sells. Mostly curiosity.
Not looking for funding. Not selling consulting. Just building things and writing about them when we remember to.
GenAI apps, experiments, and demos. Some are useful. Some are "useful." All of them taught us something — usually that we should have gone to bed earlier.
Private agents, private inference
A privacy-preserving runtime for agents and models: default-deny egress, hard tenant isolation, and inference that never needs a public API. Built for teams who will not ship prompts to someone else's endpoint — DPA or not.
"Working PoC, not a finished SaaS. Isolation is the product. Diagrams are the easy part."
Make your Kindle shine (allegedly)
Discover free classic books tailored to your taste and send them to your Kindle in a snap. Reading is good. Remembering your password is harder.
"We built an app for people who love reading. Ironically, we mostly read API docs."
Bedtime stories without the writer's block
Personalized, safe stories for children aged 4–7 with AI narration. Purple elephants optional. Parental sanity not guaranteed.
"Every story is unique. So is every typo the AI confidently insists is fine."
Your emails, now in voice
Turn your last 24 hours of Gmail into a 2–5 minute voice summary. For when inbox zero is a fantasy but inbox audio is achievable.
"Three free demos per email. We are not a SaaS. We are barely a demo."
A calculator so slow it makes you question reality
An experiment in AI-centric UIs: the LLM designs the interface via tool calls. Yes, it uses an API call per keypress. No, we are not sorry.
"Who in their right mind would use an LLM to run a calculator? Us. That's who."
Longer thoughts on AI, agents, and the occasional cubic activation function. We try to be honest about what we don't know. Which is most things.
Sovereign environments. AI agents that never need the open internet—or someone else’s model API.
Chat won't be the future UI. Here's an experiment in AI-centric interfaces — a calculator so slow it makes you question your life choices.
Building Whispery taught us that demos are a feature, not a bug. Three free tries and we're done — privacy included.
Got a weird AI idea? Want to chat about agents, scaffolding, or why ReLU gets all the attention? We're friendly. Occasionally helpful. Never corporate.
Consulting is not our main thing, but if your problem is interesting enough we'll probably reply anyway.