Tuesday, August 25, 2026
Your Lumis briefing — Tuesday, August 25, 2026
#1
Study: LLM leaderboards are artifacts of harness config, not model quality
Researchers citing multiple-choice benchmark rankings should re-check option ordering/prompt wording sensitivity before trusting leaderboard deltas as real capability gaps.
→ arXiv cs.AI
#2
NVIDIA absorbs Poolside in $12B 'reverse execuhire,' founders retain $1B stake
Signals NVIDIA vertically integrating model/infra talent directly; watch for shifts in coding-model competition and neocloud capacity commitments (7GW Infraco scale-up).
→ Latent Space
#3
Windows MS Paint and Photos secretly embed GUID watermarks in AI-generated output, even offline
Anyone building local/offline AI image pipelines on Windows should audit outputs for hidden identifiers before assuming air-gapped tools are watermark-free.
→ Hacker NewsGet it in your inbox
Want this every morning,
tailored to your sources?
Lumis synthesizes Hacker News, arXiv, The Batch, Latent Space — and any RSS feed you choose — into three sharp items before your day starts.
No spam. Invite when your slot opens.
You're on the list. We'll be in touch.