giordanol

Bamboo growth pattern - years of invisible growth underground, then it suddenly shoots up 90 feet.

Easy to forget how long the roots were forming.

Publishing early work feels pointless until you look back and realise the later stuff couldn't exist without it. Same goes for any expressive work. Sounds like a platitude, but it really is all about the process.

The tooling problem is 90% solved. The new technical bottleneck is human judgment.

Feels like a semi-simple UX fix could make this a lot more natural. Git-style forks but for chats.

Would love to see metrics that isolate recovery behaviour (if any)

Agree with this. Constraining generation with physics, legality, or even tooling limits turns the model into a search-and-validate engine instead of a word predictor. Closer to program synthesis.

The real value is upstream: defining a problem space so well that the model is boxed into generating something usable.

Shift feels real. LLMs don't replace devs, but they do compress the value curve. The top 10% get even more leverage, and the bottom 50% become harder to justify.

What worries me isn't layoffs but that entry-level roles become rare, and juniors stop building real intuition because the LLM handles all the hard thinking.

You get surface-level productivity but long-term skill rot.

Cursor’s doc indexing is acc one of the few AI coding features that feels like it saves time. Embedding full doc sites, deduping nav/header junk, then letting me reference @docs inline actually improves context grounding instead of guessing APIs.

LLMs shift the bottleneck - becomes less about typing code, more about spotting when something’s subtly wrong. Still need real judgment just applied to different layers. The skills that atrophy are surface-level. The deeper ones (debugging, systems thinking, knowing what not to trust) become more important.

Don't think the limit is in what LLMs can evaluate - given the right context, they’re good at assessing quality. The problem is what actually gets retrieved and surfaced in the first place. If the upstream search doesn’t rank high-quality or relevant material well, LLM never sees it. It's not a judgment problem, more of a selection problem.

Basically treating extraction as an adaptive loop instead of a static function. If first parse fails or looks incomplete, tweak the prompt, inject more context, or switch strategies. Memory helps carry forward partial wins so you don’t start from scratch. We’ve seen the same pattern in agentic web environments. Structured retries, context propagation, and memory turn brittle flows into robust automation, especially with high-variance input and fuzzy schemas.