Reading up on GPT-5.6 Luna
5 deep · digging since aug 19
- Cheaper LLM labelling
Describes a Perl script that uses a cheap LLM to label git commits, then speeds it up with a streaming logistic‑regression classifier and label caching.
- The plunging price of thought
Epoch AI reports that the cost to achieve a given level of AI performance has fallen roughly 47% per quarter since 2023, outpacing historic tech price drops.
- How we closed 1,500 GitHub issues in one month
The Next.js team deployed an AI agent called closability to investigate and close over 1,400 stale GitHub issues, reducing the backlog below 1,000 while preserving relevant reports.
- The machine never raises its voice
LLMs consistently prefer their own machine-generated literary passages over human-authored classics, valuing precision and restraint over voice and strangeness when evaluating literary quality.
- Do All Your Agents Really Need Models Like Claude 5 or GPT-5.6?
Many AI agent tasks don't need frontier models; matching model capability to task complexity can cut costs by up to 75%.