From the lab
Written in the open.
No ghost-writing.
Technical notes, opinions, and research updates from the people building Born. Everything is tied to real work in the repo, not anonymous filler.
Born-9B Preview is finally here
The launch note for Born-9B Preview: Repath at the controls, GPT-5.5 Codex managing the run, the datasets, failures, benchmarks, and the public adapter.
v1
Research
Born Chat is open: PromptKit UI, live model lanes, and the public training loop
What shipped in Born Chat, how the Born and comparison lanes are organized, why the training notice is explicit, and what Born Chat Pro is meant to become.
Why Codex GPT-5.5 Operated Born-9B
The operating model behind the run: Codex GPT-5.5 High, Repath Ray Khan at the controls, and a model release managed through evidence, caveats, and repository state.
How Born-9B Learned to Breathe
The origin story of Born-9B told from the real notes: the cheap first run, the honest tie, the many teachers, and the million-token second inhale.
Born-9B v1: Distillation at the edge of 9 billion parameters
How we built a competitive coding-agent model in public: the data, the teachers, and the honest eval results. What worked, what did not, and every number we tracked.
Building the Ring teacher mix: six models, one student
The data pipeline, teacher selection criteria, and quality filters behind Born-9B's training corpus. Why we chose six teachers and how their outputs were weighted.
Why evaluation-first is the only honest way to ship AI
Benchmark theater is easy. Evidence you can trust is hard. Here is how Born thinks about measurement, and why we publish evals before we publish claims.
QLoRA on a single RTX 6000 Ada: what is actually possible
A practical account of running QLoRA fine-tuning on 48 GB VRAM. Memory budgets, gradient checkpointing, batch size tradeoffs, and measured throughput.
Dataset provenance: why every training row needs a paper trail
We document where every row in Born-9B's dataset came from. This post explains the tooling, the YAML manifests, and why provenance matters more than row count.