perrygeo

11 ms·

151.
by perrygeo 8mo ago
> The features I am currently working on cannot be vibe-coded, because no AI can understand the context. It's always possible to improve understanding and put the context into writing, usually just that no one has. Clean up the vari
152.
by perrygeo 8mo ago
Witness the giant leap forward in the capabilities of coding agents over the last year. There has been no such leap in LLM model performance. I think the causality is crystal clear. It's nothing about "AGI" and all about exis
153.
by perrygeo 8mo ago
I must use AI differently than y'all. Do we not use plan mode? There is almost no value in watching the stream of intermediate tokens. There's no need to micromanage the agent's steps. Just monitor the artifact and insist the
154.
by perrygeo 8mo ago
For me it comes down to Language. They're LLMs after all. They pattern match on tokens, and if your tokens have muddled semantics, you've lost before you even started. I have a codebase where variables are named poorly - nah that&
155.
by perrygeo 8mo ago
This. Historically there's been a lot of resistance to the idea of refactoring or refining features. The classic "It works, just ship it" mentality that leaves mountains of tech debt in its wake. And there _was_ a good reason
156.
by perrygeo 8mo ago
I feel the same way. LLMs errors sound most plausible to those who know least. On complex topics where I know what I'm talking about, model output contains so much garbage with incorrect assumptions. But complex topics where I'm o
157.
by perrygeo 9mo ago
Having been in this situation more than once, recreating a concept from scratch when you've already coded it once takes ~20% of the time. This also tracks with my long term empirical observation that roughly 80% of a software project i
158.
by perrygeo 9mo ago
> (configure to run in CI) Every time I've believed this, I came back to broken CI. Depending on your language ecosystem, your CI workflows and build deps stand a greater chance to break than your code.
159.
by perrygeo 9mo ago
I jumped into a new-to-me Typescript application and asked Claude to build a thing, in vague terms matching my own uncertainty and unfamiliarity. The result was similarly vague garbage. Three shots and I threw them all away. Then I watched
160.
by perrygeo 9mo ago
Thank you. Fixing a vitamin deficiency is an obvious benefit to physical health. But does it "cure depression"? To the extent that your depression was caused by vitamin deficiency, sure. But then didn't you just cure a vitami
161.
by perrygeo 9mo ago
Wat? Android, Linux, BSD (incl iOS and Mac) absolutely dominate the market in terms of mind share and deployment. I can't think of another philosophy that is so directly, causally responsible for billions of dollars of realized value.
162.
by perrygeo 9mo ago
I guess I will never understand the microservices vs monolith debate. What about just "services"? There are 1001 reasons you might want to peel off functionality into a separate service. Just do that, without making it into some p
163.
by perrygeo 9mo ago
I get that Gas Town is part tongue-in-cheek, a strawman to move the conversation on Agentic AI forward. And for that I give it credit. But I think there's a real missed opportunity here. I don't think it goes far enough. Who wants
164.
by perrygeo 9mo ago
Speaking only of written communication here: I've noticed a distinct trend of people stopping documentation, comments, release notes, etc. intended for human consumption and devoting their writing efforts to building skills, prompts, C
165.
by perrygeo 9mo ago
For me, the loneliest period of my life was when I was socially active but hanging out with people that I didn't really like or respect. Don't neglect spiritual and mental health as a strong component of loneliness. It's not
166.
by perrygeo 9mo ago
It's remarkable how these papers show a deep understanding of programming 50 years ago. Even with anemic hardware, the limit is always in the programmers brain - as uncomfortable as that is to admit. Half a century of new tech and AI a
167.
by perrygeo 9mo ago
I don't know what it is about observability that brings out the over-engineering in us. I haven't actually measured it but I suspect many startups doing things the "modern" way (ie logs, distributed traces, metrics, ci&#
168.
by perrygeo 9mo ago
In a language that is otherwise as simple as it could possibly get away with (no `if`!), `use <-` initially feels like magic and somewhat out of place. But take look at nested callback code, the pyramid of doom, and you see why it's
169.
by perrygeo 9mo ago
LLMs tend to rise to the level of the complexity of the codebase. They are probabilistic pattern matching machines, after all. It's rare to have a 15 year old repo without significant complexity; is it possible that the reason LLMs ha
170.
by perrygeo 9mo ago
I've never thought of open source as something you can make money on directly. It's hard to see how it benefits an IC economically, besides getting some recognition and a sense of pride. Open source has always felt explicitly like
171.
by perrygeo 9mo ago
This. Mammals are (generally) K-selected species, meaning they invest heavily in raising their young. In the absence of natural pressures, mammals reproduce like crazy until they bump up against the environment's carrying capacity. Hum
172.
by perrygeo 9mo ago
Even with git-prime reducing the address space by a few orders of magnitude, there's still (effectively) zero chance for collision. The difference between 10^-29 and 10^-27 isn't that great in practice.
173.
by perrygeo 9mo ago
50x more code? Absolutely plausible. 50x more ideas implemented, or 50x better ideas? Doubtful. Generating code doesn't generate value.
174.
by perrygeo 9mo ago
> low fidelity version of a browser UI? That's the point. For me, with very few exceptions, modern web UI is steaming pile of dogshit - no consideration for user's attention, speed, or usability. TUI are extremely low fidelity;
175.
by perrygeo 10mo ago
This is exactly what I mean. You cite a single 12-yr old study and extrapolate its conclusion to extremes. No mention of the Swedish cohort study (Lindqvist 2016) showing sun-avoiders had 2x mortality risk over 20 years. No mention of the d
176.
by perrygeo 10mo ago
As with anything health/nutrition related, the debate gets so comically two-sided with reductionist arguments talking past each other. Get some sunlight but don't get a sunburn. It's not rocket science.
177.
by perrygeo 10mo ago
Don't think of nix-darwin as an alternative to homebrew. Por que no los dos? You can run homebrew (declaratively!) via nix-darwin, and I think they complement each other quite well. I stick to nixpkgs for most of the foundational sys
178.
by perrygeo 10mo ago
This is evidently a controversial take: if your projects are always delayed, maybe systematically track the things that slow you down and fix them? In practice, this means instead of working around tech debt, you pay it down prior to workin
179.
by perrygeo 10mo ago
For a show that takes ample time to over-explain every detail of the plot through repetition and analogy, I can't see how anyone would be confused by it. It's all there in the script a dozen times over. I thought is was a good end
180.
by perrygeo 10mo ago
> small int (0-256) cached It's -5 to 256, and these have very tricky behavior for programmers that confuse identity and equality. >>> a = -5 >>> b = -5 >>> a is b True >>> a = -6 &g

More ›