Hacker News: waldrews

New comment by waldrews in "Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies?"

waldrews — Fri, 23 Oct 2026 16:55:47 +0000

The separate lines of business need much smaller human controlling teams now, but they still need a few responsible humans each, that will make sure the separate thread never creates a problem that bubbles up to become a drain on OpenAI's leadership's time and cognitive load, which should be reserved for their critical path. It's far less risky if those secondary businesses that aren't strategic to 'make and control the leading AI platform' are not OpenAI's responsibility.

New comment by waldrews in "Advanced coding in low-resource programming languages"

waldrews — Tue, 20 Oct 2026 18:50:49 +0000

Soon enough, it'll be 'Could we not entrust our safety critical systems to those emotional error prone humans? Pretty Please?'

New comment by waldrews in "A browser-native classic Visual Basic VB6 IDE"

waldrews — Mon, 05 Oct 2026 05:41:52 +0000

The property grid control is still the greatest general purpose UI of all time. Define your classes with some attributes and it just works. It was all downhill from there with web and mobile.

New comment by waldrews in "Singapore govt dating app uses Gale-Shapley stable marriage algorithm"

waldrews — Wed, 30 Sep 2026 22:41:22 +0000

It's about prioritizing candidates for the costly initial audition, and breaking symmetry for the introduction, something particularly useful for people who can't just approach other at random in a social setting.

New comment by waldrews in "Gemini 4 Argon"

waldrews — Wed, 30 Sep 2026 20:46:45 +0000

Dear Google, please don't turn off your old generally available Pro-class model before your new Pro-class model is generally available (previous discussion https://news.ycombinator.com/item?id=49668196 )

New comment by waldrews in "Ask HN: Did Google kill its enterprise workhorse model?"

waldrews — Mon, 14 Sep 2026 18:16:31 +0000

That's fine if you're doing interactive dev tasks, but we're in the large volume, cost effective, big inputs, business still with low error tolerance business, and tuned the heck out of what we can get with minimal fix cycles. Millions of cases at hundreds of thousands tokens each - after all the prefiltering by cheaper models - and the tasks still need them to do convoluted reasoning. So 'usually get it right the first time' is a big part of the cost equation.

New comment by waldrews in "Ask HN: Did Google kill its enterprise workhorse model?"

waldrews — Sat, 12 Sep 2026 22:51:07 +0000

Yup. The problem is that it's bizarrely still not in General Availability status.

New comment by waldrews in "A few good ideas in programming languages"

waldrews — Sat, 12 Sep 2026 18:45:56 +0000

Completely free flow typing is risky in terms of interpretability, but type narrowing - var a : supertype; if (a is subtype) { // a is known to be subtype }, or type case, saves boilerplate in any OOP language.

New comment by waldrews in "Ask HN: Did Google kill its enterprise workhorse model?"

waldrews — Sat, 12 Sep 2026 18:22:49 +0000

Yup, tried all variations of that. There's the advantage that you can use a lower hallucination OCR specific model for the pre-processing, at least for clean text. But for something hard like handwritten forms, applying VLM with context is less error prone than preprocessing to text.

Also - and this is bizarre - the token cost of doing that is higher, not lower, at least in Gemini world, and by a large margin. That's very counterintuitive, but a page encoded as image tokens can be smaller than same page as text, and is not meaningfully lossy on documents that are just typed text because the models are well trained on those.

New comment by waldrews in "Ask HN: Did Google kill its enterprise workhorse model?"

waldrews — Sat, 12 Sep 2026 18:10:12 +0000

We sure did. It's a great writer, better in a harness, will process lots large context, but complex reasoning with convoluted rules and low hallucination tolerance? That's still larger model territory.

New comment by waldrews in "Ask HN: Did Google kill its enterprise workhorse model?"

waldrews — Sat, 12 Sep 2026 05:13:02 +0000

It's still 'preview' and not generally available, so can't run it for US restricted workloads.

Ask HN: Did Google kill its enterprise workhorse model?

waldrews — Sat, 12 Sep 2026 02:41:49 +0000

Is anyone else in a panic over the Gemini 2.5 model generation (Pro, Flash) being sunset in October before there's even any Pro class model in general availability (with geo restrictions etc.)? Google wants everyone to migrate to 3.x Flash, which beats the older Pro models on the benchmarked tasks, but isn't the same thing as the Pro class on reasoning-heavy tasks like complex reasoning on very large documents (my big use case).

The Gemini family had a distinct niche in document comprehension, with thousand page input documents taking only 300k tokens. Nothing quite like that in OpenAI or Anthropic world, even at more than 10x the token adjusted price. Should we just give up on Google at this point and engineer around the competitors' limits and eat the costs? Totally unnecessary own goal by team Google.

Comments URL: https://news.ycombinator.com/item?id=49668196

Points: 27

# Comments: 18

New comment by waldrews in "The creator of Jujutsu has joined ERSC"

waldrews — Tue, 01 Sep 2026 18:55:53 +0000

(off topic) why the headache-inducing animated background? An annoyance for all, and an actual accessibility issue for some.

New comment by waldrews in "Show HN: How much of Hacker News is about AI?"

waldrews — Wed, 26 Aug 2026 17:15:41 +0000

Every technical topic involves AI now. Even if it's about why the ancient Greeks didn't turn their steam engines to industrial use - the AI's should be researching the details. We might as well ask which topics involve humans or use language.

New comment by waldrews in "TreasuryDirect: Prepare for ID.me – Your New Way to Log In"

waldrews — Sat, 15 Aug 2026 17:50:28 +0000

TreasuryDirect's login and account recovery experience has been notorious for years, both for user experience and for people easily getting locked out for weeks. It's good they're being careful with this rollout, as it serves both individual and institutional accounts where dollar amounts involved are epic even by bank standards, and rarely checked by hand, so even single account breaches are serious.

New comment by waldrews in "Mistral OCR 4.1"

waldrews — Thu, 13 Aug 2026 21:53:12 +0000

The VLM's are so good at complex document understanding now. But you just can't trust them not to invisibly censor sensitive clinical/legal docs, even at the maximally permissive settings.

And the deep learning OCR-only models won't censor, but can and do hallucinate. I've yet to see a 'scan with different approaches and reconcile and say you're not sure if they don't agree' system just work for generic complex documents.

New comment by waldrews in "Squeak 6.1"

waldrews — Mon, 10 Aug 2026 17:37:31 +0000

The only mainstream-ish language where that still happens? R. Too bad, there's a lot more use cases for this sort of thing now - versioning, anything non-persistent agents touch, collaboration, auditable enterprise LOB. It's 2026, why are we (or our agents) still writing serialization code? Even if the AI's write the boilerplate, the state management fragility is often a tax/risk.

New comment by waldrews in "Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber"

waldrews — Tue, 21 Jul 2026 20:10:32 +0000

3.5 Flash-Lite seems available in US region, as was 3.5 Flash; but 3.6 Flash looks Global only so far when pinging. If Google employees are watching, will this issue go away?

New comment by waldrews in "Einstein's relativity rules chemical bonds in heavy elements, new research shows"

waldrews — Sat, 11 Jul 2026 03:08:00 +0000

Very farsighted, after working as a patent clerk, to lay claim on such a foundational technology. Back in the day, they must've been like, oh, so Mercury blocks the sun at the wrong time, but where's the commercial value - and now every chemical company throughout the universe is about to get a bill every time they make something more complex than hydrogen gas.

Meanwhile, Galilean relativity has long gone out of patent, and people on board planes and other vehicles just move around like they were in a stationary reference frame paying no royalties.

New comment by waldrews in "SpaceX to buy Cursor for $60B"

waldrews — Wed, 17 Jun 2026 03:18:11 +0000

Back in my day, a software company of national importance would be considered a big success if acquired for 60 million... Of course, at that point it would have to be profitable, and actually own its key IP...