About Bowl of Data
A weekly digest at the intersection of AI, security, blockchain, and engineering — curated by humans, powered by Maki, and run mostly on local models.
Our mission
Staying current with technology should take minutes, not hours. Every week, hundreds of articles, papers, and release notes compete for your attention — and most of it is noise. Bowl of Data reads all of it so you don’t have to, then hands you the short list that actually matters.
Our focus is deliberate: breakthroughs in AI and machine learning, exploits that reshape security, meaningful moves in blockchain, and engineering worth adopting. No hype, no filler — just the signal.
What’s in every issue
Each issue is built from three independent streams, curated on their own tracks so one never crowds out the others.
This week in tech
Stories
The strongest reads from across the web, each with a TL;DR, a longer write-up, and a link to the source.
Research
Papers
Notable work from arXiv and Hugging Face, selected on its own track for significance and novelty.
Model watch
Releases
Every new model and major update from the labs — Anthropic, OpenAI, Google, Meta, Mistral, NVIDIA, and more.
How it works
Gather
Maki, our multi-agent framework, scans hundreds of sources every week: RSS feeds, Hacker News, Reddit, GitHub, Lobste.rs, arXiv, Hugging Face, and the announcement pages of every major AI lab. Live trend signals from Google Trends, Reddit, and GitHub steer what earns a closer look.
Read & rank
Every candidate is read by a language model that extracts its topic, key points, technologies, and a quality score, then ranks the field — weighted by what the wider community is actually talking about that week.
Summarise
Each surviving story earns a two-sentence TL;DR and a three-paragraph long-form resume, so you can skim the gist or go deep without ever leaving the page.
Curate
Separate curation passes pick the best articles and the best papers, guided by eight weeks of editorial memory — so the mix stays fresh, balanced across topics, and never repeats itself.
Review & publish
The team reviews the shortlist — the last gate before anything ships. Then the same selection goes out three ways: here on the site, as a Substack digest, and as a weekly podcast with a locally synthesised voice.
Local by design
Almost all of that work runs on open models on our own hardware — not a commercial AI API. The articles we process, the prompts we send, and the summaries we write never leave our machines, and even the podcast voice is synthesised locally.
Two things follow from that. Your data and ours stays private — nothing is handed to a third-party model provider to log or train on. And with no per-token API bills, we can read far more sources, far more thoroughly, at a fraction of what the same pipeline would cost on a hosted service — savings that keep the newsletter free.
Private
Processing stays on our own hardware. No third-party AI service sits in the loop.
Low-cost
Open local models mean no per-token bills, so coverage stays broad and the newsletter stays free.
Open models
Built on open-weight models we run ourselves, with the podcast voice synthesised locally too.
Get involved with Maki
Maki is currently in private development: the repository is not yet public, as the project is still in its early stages. That said, we believe in building with the right people from the start, and collaboration is already open.
Whether you are a developer interested in contributing to the framework, a technical writer who wants to document AI agent systems, or a reviewer who can help sharpen the newsletter’s quality: send a request through the contact page and we will get back to you.
Frequently asked questions
What is Bowl of Data?
How is the newsletter curated?
Is Bowl of Data free?
How often is it published, and where can I read it?
Join the digest
Free, weekly, no spam. Just the stories that matter.
Subscribe on Substack