The rise and fall of Jev - Alex Ewerlöf Notes

kept by baldie oct 23

Jev, a fast non-reasoning classifier branded as a 'System One Model,' quickly rose and fell from hype as benchmarks showed it cannot reason and rivals offer better performance-per-cost.

Surely the crash of the US economy has to be soon
Hacker News commenters debate whether a US economic crash is imminent, citing an AI investment bubble, distorted unemployment data, and deepening wealth inequality as key warning signs.
also on #ai-bubble
AI IQ — Intelligently Measuring AI Intelligence
The piece presents a methodology that maps public benchmark scores to IQ estimates across seven dimensions, enabling comparison of AI models by intelligence, speed, and cost.
also on #ai-evaluation, DeepSeek, OpenAI
AI evals are becoming the new compute bottleneck
AI evaluation costs now rival or exceed training costs, with agent benchmarks reaching $40,000 per sweep and training-in-the-loop evals resisting effective compression.
also on #ai-evaluation
How Jev works: calibrated decision models
Jev is a low-latency decision model that scores options via LLM forward passes to output calibrated probabilities without text generation.
also on Typesafe.ai, Jev, #software-engineering
i hate the future: the normalization of inexplicable failures
Argues that AI-driven development normalizes unexplained failures, turning accountability into shrug‑off attitudes like “stupid thing sucks,” undermining proper debugging software.
also on Typesafe.ai, Jev, #software-engineering