Thread

Jev in 25 Lines of Python

691 points · 212 comments · bashbjorn

teravor · 1790181904 · focus · HN ↗
to be fair a Jev architecture would be better optimized for this particular workflow than an LLM.
```
    &lt;think&gt;\n\n&lt;&#x2F;think&gt;
```
but letting an LLM think would trade latency and performance for significant reliability above that of Jev.

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.