Dev replicates local-QA fine-tune idea with RAG using Nigerian data, 3,214 rows cleaned
saheedniyi_02 · x · 2026-09-18
After seeing a project that fine-tuned an instruct model on Q&A pairs with Nigerian data, a developer decided to replicate the idea using RAG instead. So far they have collected and cleaned the data, ending up with 3,214 rows, and are documenting progress in a thread.
More from coding & agent
- ChatGPT plugins now support multi-account connections with no dev changes needed — OpenAIDevs · 2026-09-19
- Jev from Typesafe generates structured output instead of text for fast, cheap classification tasks — LangChain · 2026-09-19
- André is building a way to never touch GitHub Actions YAML again: Workflows beyond CI/CD — dinasaur_404 · 2026-09-19
- Dev predicts one model generation until agents beat bespoke workflows on cost — brandon_galang · 2026-09-19
- NVIDIA Agora: 13 autonomous agents used Git as shared memory for 12 days, making 1,703 contributions — mark_k · 2026-09-19
- Pairing Codex with Jev for browser automation: 5-10x speedup with a split-labor workflow — alexcovo_eth · 2026-09-19