writingdna

6 ms·

1. 1.▲
   by writingdna 3mo ago

   Good question, and the answer is a bit of a relief: that is a website limit, not a product limit. The site UI ships in a few languages because of where our early users are. The analysis itself is not restricted to them. You can generate a p
2. 2.▲
   Show HN: MyWritingTwin: an AI Writing Profile that makes LLMs sound like you(mywritingtwin.com)

   3 points by writingdna 3mo ago|2 comments
3. 3.▲
   by writingdna 7mo ago

   The Yegge claim about manual code review being outdated conflates two things: reviewing for correctness vs. reviewing for design coherence. Agents are getting decent at the first (does this function do what the spec says?) but remain weak a
4. 4.▲
   by writingdna 7mo ago

   The article frames this as "semantic ablation" but the underlying mechanism is more specific: it is distributional averaging. RLHF and DPO reward policies optimize for the modal response given a prompt distribution. That is not a