Hacker News: trefoiled

New comment by trefoiled in "Vomit: Clean up Claude 5's token output with a separate LLM"

trefoiled — Thu, 20 Aug 2026 16:27:22 +0000

I've been grappling with this for weeks, not just in Claude but in Codex as well, which isn't quite as bad but still annoying. AGENTS.md does very little, agents will consistently violate the communication preferences, especially as the session drags on. It's incredible to me that there's no good way to reliably change the way an LLM responds to you that a workaround like this would even be necessary. It seems like such a failure to live up to the promises of the product.

The baked in communication style of these models is so obnoxious it's impacting my work. The best way I can describe it is that everything is optimized to impress the user and make the agent sound more authoritative, but the way this is done is through deliberate obfuscation, inserting inappropriate and extremely dense jargon, and bizarre, stilted metaphors. It's like they've been trained to produce output that's hard to read.

New comment by trefoiled in "How to stop Claude from saying load-bearing"

trefoiled — Tue, 14 Jul 2026 18:17:48 +0000

I've spent two hours today trying to provide Sol with guidance that reduces its pretentiousness, to no avail. Layers upon layers of rules only for it to use the phrase "async spline resolution" in a sentence.

New comment by trefoiled in "LastPass notifies users of yet another data breach"

trefoiled — Thu, 25 Jun 2026 16:28:30 +0000

Unfortunately it's one of the most bug-ridden and unreliable pieces of software I've ever used. I encounter issues with it on a daily basis, but the burden of switching and a lack of superior options keeps me locked in.

New comment by trefoiled in "When AI Builds Itself: Our progress toward recursive self-improvement"

trefoiled — Thu, 04 Jun 2026 20:28:42 +0000

Forget about the danger of a dev to customer pipeline with no product people in between, some of us are living with the reality of product to customer pipeline with no developers in between, and that's much more disturbing. Our CEO is now the top contributor to our codebase, and he's completely non-technical.

New comment by trefoiled in "Where's Firefox going next?"

trefoiled — Wed, 16 Jul 2025 17:04:21 +0000

> Features like tab groups, vertical tabs, profiles, new tab wallpapers, PWAs, and taskbar pinning weren’t just ideas – they were direct responses to what you told us you wanted

Yeah, that's ChatGPT. And not a particularly high quality ChatGPT style sentence. They weren't just ideas, they were direct responses? Ok.

New comment by trefoiled in "Dubious Math in Infinite Jest (2009)"

trefoiled — Tue, 10 Jun 2025 16:21:21 +0000

This isn't the only example of a debate over intentionality in mistakes in Infinite Jest. The book's french is also littered with errors so egregious that most think they could only have been intentional [1].

[1] https://ask.metafilter.com/116066/French-language-in-Infinit...

New comment by trefoiled in "Why Archers Didn't Volley Fire"

trefoiled — Mon, 05 May 2025 13:28:29 +0000

I found the overuse of italics for emphasis fatiguing

New comment by trefoiled in "Avoiding skill atrophy in the age of AI"

trefoiled — Sat, 26 Apr 2025 00:06:01 +0000

If your butcher felt the same way you did, he wouldn't exist

New comment by trefoiled in "A Scaled Down Look at Spending, Revenue, and What's Being Cut"

trefoiled — Fri, 25 Apr 2025 23:54:53 +0000

It's also slightly deceptive to describe the DOGE cuts to USAID programs as "one time cuts." This is technically true, but it implies that the programs will be replaced with different programs with a similar cost, which likely isn't the case considering the Trump admin is trying to axe USAID permanently. The actual savings are whatever USAID's budget would have been had it continued to exist.

New comment by trefoiled in "On loyalty to your employer (2018)"

trefoiled — Thu, 24 Apr 2025 16:12:39 +0000

There's an exception to this I've seen since a relative started working in the game industry. There are executives in that industry who have a retinue of loyal followers. The studios the executive works for may change regularly, but his followers come with him each time. These workers will spend their entire career serving one man, and in exchange he always has a job lined up for them and seems to trust them the same way they trust him. It's very different from my experience in the rest of the tech industry, but I'm sure it happens to a limited extent there too.

New comment by trefoiled in "Synology Lost the Plot with Hard Drive Locking Move"

trefoiled — Tue, 22 Apr 2025 17:10:33 +0000

I was disappointed when I fully understood the limitations of SHR after purchasing my Synology box, and subsequently failed to install MergerFS on it. It's the only thing I miss about my old self managed server.

New comment by trefoiled in "How the internet became shit"

trefoiled — Fri, 19 Apr 2024 18:50:39 +0000

AdGuard is system wide and works well for me, but it isn't free

New comment by trefoiled in "My productivity app is a never-ending .txt file (2022)"

trefoiled — Tue, 20 Feb 2024 23:12:11 +0000

Because this sounds so much like Logseq, I'm curious what about it didn't meet your needs

New comment by trefoiled in "How to tackle unreliability of coding assistants"

trefoiled — Wed, 29 Nov 2023 19:49:51 +0000

I hear this argument applied often when people bring up the deficiencies of AI, and I don't find it convincing. Compare an AI coding assistant to reaching out to another engineer on my team as an example. If I know this engineer, I will likely have an idea of their relative skill level, their familiarity with the problem at hand, their propensity to suggest one type of solution over another, etc. People are pretty good at developing this kind of sense because we work with other people constantly. The AI assistant, on the other hand, is very much not like a human. I have a limited capacity to understand its "thought process," and I consider myself far more technical than the average person. This makes a verification step troublesome, because I don't know what to expect.

This difference is even more stark when it comes to driving assistants. Video compilations of Teslas with FSD behaving erratically and most importantly, unpredictably, are all over the place. Experienced Tesla drivers seem to have some limited ability to predict the weaknesses of the FSD package, but the issue is that the driving assistant is so unlike a human. I've seen multiple examples of people saying "well, humans cause car crashes too," but the key difference is that I have to sit behind the wheel and deal with the fact that my driving assistant may or may not suddenly swerve into oncoming traffic. The reasons for it doing so are likely obscure to me, and this is a real problem.

New comment by trefoiled in "Fire TV now also displays full-screen video ads on its homescreen"

trefoiled — Sat, 25 Nov 2023 18:57:34 +0000

got it, thank you!

New comment by trefoiled in "Fire TV now also displays full-screen video ads on its homescreen"

trefoiled — Sat, 25 Nov 2023 16:49:03 +0000

do you mean through Airplay?

New comment by trefoiled in "Fire TV now also displays full-screen video ads on its homescreen"

trefoiled — Fri, 24 Nov 2023 19:30:57 +0000

this feels like a race to the bottom which every industry player except Apple is competing in. I never would have considered an Apple TV just a few years ago, but my Shield is gathering dust ever since I made the switch.

New comment by trefoiled in "OpenAI's employees were given two explanations for why Sam Altman was fired"

trefoiled — Tue, 21 Nov 2023 06:02:16 +0000

Weirdly plausible considering Tasha McCauley also works for the RAND Corporation

New comment by trefoiled in "Mass producing the most expensive rice cooker [video]"

trefoiled — Sun, 05 Nov 2023 15:51:18 +0000

My Zojirushi makes rice with better consistency and texture than any rice I've had outside of a sushi restaurant - but then again it's possible I just don't know anyone who knows how to make good rice on a stove

New comment by trefoiled in "Helix 23.10 Highlights"

trefoiled — Thu, 26 Oct 2023 15:12:57 +0000

this is the big problem. I hope there's a fix for it someday.