Hacker News · upvotes history

LLMs reward expertise

SteadyVane has taken 40 readings of LLMs reward expertise since Aug 3, 2026: 174 upvotes at the first reading, 1,409 at the latest (+1,235). Its biggest one-day gain was +1,120 on Aug 4, 2026.

open the story · Hacker News thread

174

first seen upvotes

1,409

peak upvotes

1,409

latest upvotes

observations

first seen Aug 3, 2026 · last seen Aug 8, 2026 · published Aug 3, 2026

Day by day

Day (UTC)	upvotes	Change
Aug 8, 2026	1,409	+7
Aug 7, 2026	1,402	+9
Aug 6, 2026	1,393	+12
Aug 5, 2026	1,381	+70
Aug 4, 2026	1,311	+1,120
Aug 3, 2026	191	+17

Each day's last reading. Change: against the day before (after days with no reading: since the last day that has one); the first day shows the change within that day.

Featured on 1 day: August 4, 2026.

More rising on Hacker News

Goldman Sachs is pulling Claude Code from its bankers and traders+712159 → 871 upvotes
Tell HN: Gemini trains on chat history, even for subscribers (personal accounts)+69143 → 734 upvotes
Ask HN: What's your current AI agentic setup?+5644 → 568 upvotes
OpenAI seems to be silently re-enabling model training on Codex chats+5286 → 534 upvotes
Gemini Omni 1.2 Flash+236120 → 356 upvotes
OpenAI raises $11B at a $1.25T valuation+1818 → 189 upvotes
Basically no one is using AI agents for their DoorDash orders+13345 → 178 upvotes
AI-generated posters don’t have to be horrible+1,78453 → 1,837 upvotes

Everything rising on Hacker News →