Hacker News: kraemahz

New comment by kraemahz in "Decisions that eroded trust in Azure – by a former Azure Core engineer"

kraemahz — Thu, 02 Apr 2026 22:42:08 +0000

AWS and Google Cloud are both huge and are significantly better in UX/DX. My only experience with Azure was that it barely worked, provided very little in the way of information about why it didn't. I only have negative impressions of Azure whereas at least GC and AWS I can say my experiences are mixed.

New comment by kraemahz in "GitHub is once again down"

kraemahz — Tue, 24 Mar 2026 21:30:04 +0000

Oh, they've gone beyond micro. It's Macroslop now!

New comment by kraemahz in "AI coding is gambling"

kraemahz — Wed, 18 Mar 2026 19:45:45 +0000

You also have control over the workflow they follow and the standards you expect them to stick through, through multiple layers of context. Expecting a model to understand your workflow and standards without doing the effort of writing them down is like expecting a new hire to know them without any onboarding. Allowing bad AI code into your production pipeline is a skill issue.

New comment by kraemahz in "GPT‑5.4 Mini and Nano"

kraemahz — Tue, 17 Mar 2026 20:25:18 +0000

I find both Opus 4.6 and GPT-5.4 have weaknesses but tend to support each other. Someone described it to me jokingly as "Claude has ADHD and Codex is autistic." Claude is great at doing something until it gets done and will run for hours on a task without feedback, Codex is often the opposite: it will ask for feedback often and sometimes just stop in the middle of a task saying it's done with step 1 of 5. On the other hand, Codex is a diligent reviewer and will find even subtle bugs that Claude created in its big long-running "until its done" work mode.

New comment by kraemahz in "Comparing Google and ChatGPT"

kraemahz — Fri, 02 Dec 2022 17:21:43 +0000

Between the release of GPT-3 and GPT-3.5 there was Gopher, which raised the bar on TruthfulQA from essentially random (22.6%) in GPT-3's case to 45% for Gopher. GopherCite then brought the performance up to 80-90%. One has to assume that OpenAI is using state of the art techniques in their new model releases. That the LLMs went from choosing answers randomly to producing accurate results on a great deal of questions (they still suck at math) is missed for anyone who is not aware of the historical context that shorthanding 3.5 to 3 causes.

New comment by kraemahz in "Comparing Google and ChatGPT"

kraemahz — Thu, 01 Dec 2022 19:18:43 +0000

I'm curious why everyone keeps getting confused about this model being GPT-3 and using their past experiences with GPT-3 to justify their position. The model is not GPT-3 and and at this point GPT-3 is far behind the state of the art. OpenAI calls this model "GPT-3.5".

It is also capable of far more than relaying information, as such it is also serving the purpose of Q/A sites like Stack Overflow. You can put wrong code into it and ask for bug fixes and it will return often exactly the correct fix.

Framed as a search engine it obviously fails on some measure, framed as a research assistant it exceeds Google by leaps and bounds (which suffers greatly from adversarial SEO gumming up its results).

New comment by kraemahz in "Rsync client-side arbitrary file write vulnerability"

kraemahz — Tue, 02 Aug 2022 23:34:40 +0000

I am competent enough to read c and note that none of those changes you highlighted have functional impacts. They were largely cosmetic changes (e.g. removing an else condition by assigning the inclusive flags to a variable). I saw one change that looked like a performance improvement (an early loop exit break added).

I will also note that rsync has a test suite and running tests as "integration" style testing is more common of shell utilities than writing gobs of unit tests. I would have liked to see an additional targeted test written for the CVE in this patch though.

New comment by kraemahz in "Newly Measured Particle Seems Heavy Enough to Break Known Physics"

kraemahz — Fri, 08 Apr 2022 02:13:30 +0000

I've seen other instances where physicists working at CERN would blind themselves to the results of their computations while developing their algorithms, making sure the code was bug-free and physically accurate without running it "in production" as it were on the actual experimental results. This is to blind themselves to the actual statistics of the experiment so that "p-hacking" or manipulation of the results to fit a preconceived agenda is not a temptation.

New comment by kraemahz in "What is money, anyway?"

kraemahz — Mon, 28 Mar 2022 04:10:31 +0000

I didn't read the article, but I scrolled through it until I found the place where he answers the question in the headline:

What is money?

Well, the answer to that question ties into the difference between currency and money. Currency is some other entity’s liability, and they can choose whether or not to honor that particular liability. Money is something that is intrinsically valuable in its own right to other entities, and that has no counterparty risk if you custody it yourself (although it may have pricing risk related to supply and demand). In other words, Russia’s gold is money; their FX reserves are currency. The same is true for other countries.

New comment by kraemahz in "I Can’t See You but I’m Not Blind"

kraemahz — Tue, 14 Dec 2021 17:05:33 +0000

Hypophantasia is what limited visualization is called. As with all things biological it is a spectrum of experiences.

New comment by kraemahz in "LinkedIn sued over allegation it secretly reads Apple users' clipboard content"

kraemahz — Sat, 11 Jul 2020 23:20:32 +0000

That got me far enough to get started, thanks!

New comment by kraemahz in "LinkedIn sued over allegation it secretly reads Apple users' clipboard content"

kraemahz — Sat, 11 Jul 2020 17:51:13 +0000

Is there a list of these somewhere? What are the steps needed to verify for myself? Exfiltration of data from my phone is theft; I prefer not to be stolen from.

So far I've found: Spotify Tinder Yelp Duolingo

New comment by kraemahz in "What computer and software is used by the Falcon 9? (2015)"

kraemahz — Mon, 01 Jun 2020 04:49:59 +0000

They are worried about bugs, but they address those in other ways. This is some of the most extensively integration tested software there is (I was part of the team that built the integration testing). Every commit is ran through a custom CI system that successively tests pieces of the system working together up to and including automated hardware tests. The number of lines of code dedicated to testing the vehicle is definitely much greater than the lines of code on the vehicle.

New comment by kraemahz in "NASA fixes Mars lander by telling it to hit itself with a shovel"

kraemahz — Thu, 19 Mar 2020 14:40:35 +0000

The robots can be iterated on Earth and the environments aren't much different. The problem here is NASA.

NASA is a politicized bureaucracy that has its hands tied behind its back. It cannot afford to take risks (both from a budget point of view and a politics point of view; failed missions are a wound to national pride). It cannot afford to hire the best and brightest because government payroll levels can't match industry wages.

NASA is mired in old engineering superstition and has wrapped all progress in procedures that make it impossible to iterate quickly. They've made hardware and software decisions produced for them by niche industries that charge 10-20x consumer prices for inferior hardware that is certified by NASA standards.

NASA is not the organization that made it to the Moon anymore.