A living map — updated as the field moves

The stack worth knowing — open-source, free, and changing what one person can build.

You don't need a lab or a budget to build with AI anymore. The best models run on a laptop, the best coding tools are free, and the licences let you use almost all of it commercially. This is the map I wish I'd had starting out — every tile says what it is, why it matters, and the licence, because knowing what you're allowed to do with something is half the lesson.

Open models
Run for $0
Coding tools
Mostly free
Robots & frontier
Going open
For
Anyone curious

How to read a licence (the 60-second version)

Open source isn't one thing — the licence decides what you can build on top. Four you'll see everywhere:

MIT / Apache-2.0 — do almost anything: use it, change it, sell it, keep your changes private. The friendliest licences.
AGPL-3.0 — free to self-host, but if you offer it as a service to others you must share your modifications. Common for serious tools.
Open weights — the model is free to download and run, sometimes with a use-policy (e.g. very large companies need a separate deal).
CC-BY-NC — free for research and personal use, not commercial. Read before you build a business on it.
01 / Open-source LLMs

Models you can download and run for free.

The frontier of open models is months, not years, behind the closed ones — and free. These are the families that matter.
Meta · all-rounder

Llama

Meta's family — the one that started the open-weights wave. Sizes from runs-on-a-phone to runs-a-cluster. The safe default for general work.

Open weights llama.com
DeepSeek · reasoning

DeepSeek (V3 · R1)

The one that shocked everyone — frontier-class reasoning at a fraction of the training cost, weights released MIT. Proof open can catch up fast.

MIT github.com/deepseek-ai
Alibaba · multilingual

Qwen

Alibaba's family — strong at code, maths and non-English. Most sizes are Apache-2.0, so you can ship commercially with no asterisks.

Apache-2.0 qwen.ai
Mistral · efficient

Mistral & Mixtral

French lab, famous for small models that punch far above their weight. Mixtral pioneered cheap “mixture-of-experts” for the open world.

Apache-2.0 mistral.ai
Google · on-device

Gemma

Google's open siblings to Gemini. Small, fast, genuinely good — built to run on your own hardware, including phones and single GPUs.

Open weights ai.google.dev/gemma
Microsoft · tiny

Phi

Microsoft's “small but mighty” line — trained on textbook-quality data so a tiny model reasons like a much bigger one. MIT-licensed.

MIT huggingface.co/microsoft
The library · everything

Hugging Face

The GitHub of AI models — a million-plus models, datasets and demos in one place. Where every model above actually lives. Start here.

Hub huggingface.co
OpenAI · speech

Whisper

Speech-to-text that just works, in ~100 languages, fully open. It's the engine inside my own murmur app — your voice never leaves the machine.

MIT github.com/openai/whisper
// new ones land monthly
02 / Run it yourself

Get a model running on your own machine.

A model is just a file. These tools turn that file into something you can chat with — no cloud, no API bill, no data leaving your laptop.
03 / AI coding tools

The tools that change what one person can ship.

This is the part that's genuinely new. A single person with these can now build what used to take a team. Most are free or have a real free tier.
04 / Agents & the glue

Frameworks that make models do things, not just talk.

A chatbot answers. An agent acts — calls tools, browses, runs code, remembers. These are the building blocks, plus the standard that connects them all.
05 / Token & memory tools

Make your AI cheaper, faster, and able to remember.

The unglamorous tools that decide whether AI work is affordable. Cut wasted tokens, give an agent a real memory, point it only at what matters. These are the ones I lean on hardest.
06 / The frontier

Where it's going next — from robots to physical AI.

The same wave that hit software is reaching the physical world. Some of this is already open and hackable; some is just worth watching. Both matter if you want to see around the corner.
07 / My forks

The bets I keep on my own shelf.

When I find something brilliant I fork it — so it survives, so I can read the code, so it's mine to learn from. A few more beyond the ones above.

Pick one. Build something tiny this week.

You don't learn this by reading the list — you learn it by running ollama run llama3 once, or letting an agent fix one bug. Start small. The rest follows.

Want help building? Book a call Browse my forks ↗