Hacker News: moezd

New comment by moezd in "Nobody ever gets credit for fixing problems that never happened (2001) [pdf]"

moezd — Fri, 12 Jun 2026 05:18:44 +0000

Managers will let you get away with anything if you time your reports correctly. They also don't want to sit in meetings where they are reminded of better outsourcing alternatives and they chose to dogfeed instead.

We've become too comfortable, since actual toil is no longer seen in the company: Manufacturing is overseas, customer support is overseas, logistics is an afterthought with established guarantees. Thus we want the mild weather and smooth meetings. If your engineering team is too smooth, maybe you should already branch out to help other related but "struggling" teams to get your hands dirty and noticed.

New comment by moezd in "Why SpaceX 2040 Revenue FCST $4.3T in highly unlikely"

moezd — Thu, 11 Jun 2026 04:26:22 +0000

Tesla grew from two dudes making a sports car in their garage to being a decent EV company, despite many controversies. SpaceX is already propped up with government contracts to ship payloads to ISS. What could potentially justify 40% or so sustained growth of SpaceX? Will they create the asteroid mining industry from scratch (and make every precious metal on Earth worthless... hold on), or will they carry people to multiple planets in Solar System (so what, Mars will stay less habitable than Antarctica until the Sun blows up... hold on), or will they cover the whole world with Starlink satellites, data centers etc (which they are already doing a decent job... hold on)?

What's the secret sauce? What can Musk possibly promise this time around? More memes? He already smoked crack on YouTube, what comes next and why is this worthy of pension funds to invest into?

New comment by moezd in "Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable"

moezd — Thu, 11 Jun 2026 04:16:57 +0000

Maybe off-topic, but I'm also not happy about how they butchered my boy Opus 4.6. The model that could now hallucinates regularly.

Fable isn't even that great, not to mention it drinks token by the gallon for breakfast and keeps your data hostage for 30 days.

New comment by moezd in "Nobody needs AI to search the Internet, court says in ruling against Google"

moezd — Wed, 10 Jun 2026 21:45:26 +0000

The problem with WWW was that in a graph, you needed to find a way to rank the nodes. That was PageRank and it was the right tool for the right problem. When AI tries to do that, it essentially consumes a "curated list" of pages and sounds knowledgeable to fool the ordinary man. That's not search, it's just more of the "curated content" that we're forced to consume every day.

New comment by moezd in "If Claude Fable stops helping you, you'll never know"

moezd — Wed, 10 Jun 2026 14:23:16 +0000

10 engineers build a billion dollar company and immediately realise they need 1000 people to run it. Getting 1000 Claude licenses will still be expensive, let alone maintaining those swarms of agents.

New comment by moezd in "AWS Bedrock to require sharing data with Anthropic for Mythos and future models"

moezd — Wed, 10 Jun 2026 14:03:07 +0000

That's it. If you have confidential data that you're running with Fable, you're giving that away for free. Maybe you have always been, but now they explicitly ask for it.

New comment by moezd in "If Claude Fable stops helping you, you'll never know"

moezd — Wed, 10 Jun 2026 04:09:25 +0000

Wait until it flags duplicate code as a reason to stop, then a library owner could halt code generation entirely, and then another library owner could ask to be prioritised in the selection phase. Infinite money glitch, and you only get to use code that's endorsed by Claude today (subject to change tomorrow, or 5 minutes, so say goodbye to your evals), not the most performant or making the most sense in your refactoring.

New comment by moezd in "Apple reveals new AI architecture built around Google Gemini models"

moezd — Tue, 09 Jun 2026 05:46:47 +0000

OpenAI and Anthropic, despite bombastic media coverage, are still frontier labs. Google won't go down under if Gemini doesn't sell enough anymore. Apple and Google are planning for the aftermath of the AI IPO craze, one way or another.

New comment by moezd in "Ask HN: What was your "oh shit" moment with GenAI?"

moezd — Sun, 07 Jun 2026 14:28:31 +0000

I was working on a project where the code was erroring hard, inconsistent, immensely bloated and probably written by a bit of 2023 era AI assistance. It didn't have a lot of tests, too. I had no intention to be a garbage man on that specific project, and yet fates had their own plans for me.

So I immediately went for the package.json, attached whatever test commands that could be done, then wrote some hand holding markdowns for the agents to consume. First I let it explore the codebase and take notes, then I asked for a different approach to implement, which essentially turned that "paint the sides and ship" project to a full rewrite. But guess what? All features, one by one, were implemented and extensively tested by AI. I tested all outputs myself, especially when agent got impatient and thought something was wrong in the shell env, and after a few PRs, I started to read less and only skimmed the results to see if the code "feels like it belongs with the rest". Bloat was reduced by about 25% and now all features look like they were created by a rather competent offshore software shop. Not bad, since I was also juggling a couple more projects and involvements at the same time which would've led me to certain burnout.

New comment by moezd in "Ask HN: What is your (AI) dev tech stack / workflow?"

moezd — Fri, 05 Jun 2026 18:03:11 +0000

1) Slow code. Let the agent(s) discover and plan, then launch the swarm on the confirmed implementation steps. 2) Use LSP. If nothing works, usually you can connect it via MCP. I think all coding agents support this by now. 3) Add hooks if you want to stop the coding agent from doing something nasty, or hallucinate and give incomplete output. TDD and any verification tool you can think of are your friends. 4) Skills have been a bit of hit and miss for me, especially with less capable models. So are plugins. If you know how they work, please explain to me.

That way the model doesn't go about "let me grep this specific pattern across a million files again and again" loop and burn your entire weekly budget by Monday at noon.

I'm also curious if anyone has done something cool with memory and context management that doesn't require a custom llama.cpp implementation. I also don't have the heart to let the swarm do it end to end, because LLM generated code with less capable models really does smell, no amount of spec driven or Claude.md filled style guidelines seem to fix it.

New comment by moezd in "What if remote working, not AI, is to blame for weak junior hiring?"

moezd — Mon, 01 Jun 2026 04:06:34 +0000

Anything to avoid Peter principle at all costs, I guess. I will spell it out: We have an aging middle manager problem in IT. Those with less desire to catch up on things have mandate and budget to pursue projects with sunken cost fallacies, they overestimate the cost of architectural change and they don't take criticism well.

That's what's causing junior hiring to fail, because they don't want to hire juniors with passion, because especially with AI assisted tooling, these juniors suddenly seem a lot more capable than those aging guys who need their third meeting before lunch. Thus they hire the fresh graduate as an impressionable and yet unreliable junior instead.

New comment by moezd in "Who wins and who loses in prediction markets? Evidence from Polymarket"

moezd — Fri, 22 May 2026 06:01:29 +0000

So the market is still being built: Whales come in with educated guesses, use limit orders like responsible adults and take cash home. Daredevils, amateurs and gambling addicts come in, go all-in and try to time the market, which probably never works. %1 -> 76.5% is huge, in terms of Gini coefficient it's like living in Brazil, right?

New comment by moezd in "Tell HN: Google banned Railway's account. Everything down"

moezd — Wed, 20 May 2026 13:09:44 +0000

Exactly. Whatever happened to cloud just being "someone else's computer" and "cheap scaling on speed dial"?

New comment by moezd in "Linux security mailing list 'almost unmanageable'"

moezd — Mon, 18 May 2026 14:12:45 +0000

I think it's time the report-only intake should stop. If a reporter can't reproduce at least one use case or can't summarise it in two sentences, it should be classified as spam. LLMs write beautiful reports, it's just that sometimes it doesn't bear anything resembling the truth.

New comment by moezd in "It is time to give up the dualism introduced by the debate on consciousness"

moezd — Mon, 18 May 2026 07:46:41 +0000

No, we don't even need that. When we realise that we all project consciousness claims on each other from what we observe as zeitgeist now, just to do credit assignment, most of our circular debates will disappear. But this won't happen since many powerful entities in the world ride on the moral ambiguity, and this will hold them accountable.

New comment by moezd in "Prepare for an AI Jobs Apocalypse"

moezd — Sat, 16 May 2026 14:11:27 +0000

If your idea of a job is to sit pretty in meetings, agree on increasing synergies, then go write an email or fill out a form or something, yeah they will be gone.

Anything that requires you to think in constraints, orchestration and optimization should survive in one form or another. It always does.

New comment by moezd in "MacBook Neo Review: The Laptop for the Rest of Us"

moezd — Thu, 14 May 2026 04:36:37 +0000

It's essentially a low-end laptop made from iPad/iPhone parts. Supply chain for RAMs must be crazy that they had to navigate another tier below Macbook Air.

New comment by moezd in "Software engineering may no longer be a lifetime career"

moezd — Tue, 12 May 2026 04:14:51 +0000

This should be repeated ad nauseaum: Just because CNC tooling is available, you wouldn't get rid of your wood/metal workers. You give them more scoped tasks, expect them to finish sooner with more standardized output, and... you let them experiment a bit on the side. Perhaps not with industry grade tooling unless they want to pay for fancy tooling, but like DIY, and then, you lure nontechnical people in by telling them how much you can save with DIY. Of course it doesn't matter to no techies that some of the finishes they want are truly gruesome handiwork, but hey, they are a part of the DIY community now!

That's the way. Anything else shows that they don't know how modern economy works. And let's admit it, as a bunch of IT/software people here we are terrible at this.

New comment by moezd in "EU Age Control: The trojan horse for digital IDs"

moezd — Sun, 26 Apr 2026 15:55:41 +0000

Internet used to be a cyberpunk escape hatch. Now we want to get rid of any and all anonymity at all costs, to the point of waking your phone up with a face id and your internet history tied to your government ID 1-1. This is just sad.

New comment by moezd in "Ask HN: Am I getting old, or is working with AI juniors becoming a nightmare?"

moezd — Fri, 24 Apr 2026 15:58:15 +0000

Ask AI to ruthlessly reduce cognitive debt, purge unnecessarily defensive code and be extremely pragmatic about what you want to build. If an AI junior is building you Vault when you just asked for a secret rotator script, he's just showing off. Gently pull him from the clouds, since this is also within the JD of a senior engineer.