@henokcrypto It is not coming back!
I think Anthropic comes up with something new, claiming it's better than Fable-5 in every regards, but can't perform risky projects.
Kimi K2.7 is giving a tough competition.
Kimi K2.7 ranked #2 on the ErdosBench (benchmark for artificial intelligence's problem-solving capabilities)
The part that excites me is it's Chinese, and can't be banned!
Kimi K2.7 Code lands at #2 on ErdosBench, right behind Claude Fable 5 max.
Very strong result: 13/14 coverage, 5 fully accepted solves, 5 accepted partials, 0 major false partials, 0 unsafe partials. It even matches the top model on the five accepted solved/settled results.
The most interesting part is that Kimi K2.7 looks especially strong on decisive moves, scoring 4.8 there, with 4.7 on known theorem use and 4.1 on proof hygiene. The note also calls out Problem 3 as a particular strength.
It only loses the top spot mainly on coverage and slightly lower overall audit quality. Still, this is a very impressive showing for Kimi K2.7 Code.
The funniest thing about prompt engineering is that most people still talk to AI like it’s an intern.
The best users talk to it like:
- a senior engineer
- a system architect
- an operator
- an autonomous team
Huge difference in outputs.
Claude agents can now:
- deploy on schedules
- use env variables
- persist updates for long periods
At what point do we stop calling them “tools”
and start calling them AI employees?
OpenAI considering API price cuts because Anthropic is applying pressure!
The AI war just entered its MOST dangerous phase:
cheaper models
faster releases
infinite competition
zero loyalty
We’re watching intelligence become a commodity in real time.