Hacker News: whiplash451

New comment by whiplash451 in "Open source AI must win"

whiplash451 — Sat, 13 Jun 2026 06:04:37 +0000

This could be of interest to you: https://thealliance.ai/projects/tapestry

New comment by whiplash451 in "New privacy frontier: Europe eyes crackdown on smart glasses"

whiplash451 — Fri, 12 Jun 2026 18:11:47 +0000

Why are we even allowing this in Europe? These smart glasses are just plain data collection and surveillance in plain sight. When does the nightmare stop?

New comment by whiplash451 in "If Claude Fable stops helping you, you'll never know"

whiplash451 — Wed, 10 Jun 2026 13:52:51 +0000

Not if the model is part of a broader product that happens to be very useful/relevant to private companies willing to pay a lot for it (e.g. a coding agent that can do many things but won't help you build a frontier LLM model).

My intuition is that Claude and the likes are going gung-ho after this, along all the verticals that will generate money without threatening their moat.

New comment by whiplash451 in "If Claude Fable stops helping you, you'll never know"

whiplash451 — Wed, 10 Jun 2026 11:56:52 +0000

> The moat looks deep today but it's going to become more shallow every year.

Unless the frontier labs start nerfing their models, which is exactly what seems to be happening.

The counter-point to your argument is a future where less and less un-nerfed open-source frontier models exist. Sure, China/Meta might keep commoditizing their complement by releasing un-nerfed models, but these come with their own limitations too.

I am worried that the door to great open-source frontier models might be closing by the day now.

New comment by whiplash451 in "If Claude Fable stops helping you, you'll never know"

whiplash451 — Wed, 10 Jun 2026 11:54:30 +0000

> it takes O(10) people to build one of these things

To build a working prototype, sure. To operate at production scale, definitely not. The same rule would apply to WhatsApp and many other world-scale products. Turns out that, the moment you need to monetize these machines, your O(10) stops working.

New comment by whiplash451 in "Rich Sutton on AI creativity and discovery"

whiplash451 — Wed, 10 Jun 2026 11:48:48 +0000

The TRM architecture models both the problem and the solution at the same time. You might find it an interesting read.

https://arxiv.org/abs/2510.04871

New comment by whiplash451 in "Rich Sutton on AI creativity and discovery"

whiplash451 — Wed, 10 Jun 2026 06:21:58 +0000

You might be missing that those rare humans were sitting on tons of failed or somewhat useful discoveries made by more “mediocre” humans that history forgot.

New comment by whiplash451 in "I'm skeptical about efforts to revolutionize schooling"

whiplash451 — Fri, 05 Jun 2026 06:24:47 +0000

To a large extent, the onus is on the teacher to generate interest. Most teaching until uni is mostly forced upon students.

New comment by whiplash451 in "Gmail thinks I'm stupid, so I left"

whiplash451 — Wed, 03 Jun 2026 04:53:51 +0000

Gmail has become unbearably slow over the past year.

What used to be instantaneous (like, opening an email) now takes seconds.

Google, what happened to you?

New comment by whiplash451 in "Meta launches Instagram, Facebook, and WhatsApp subscriptions"

whiplash451 — Mon, 01 Jun 2026 05:13:52 +0000

> it costs around 27$ a year per user for Meta to run the business

And a lot of this is probably spent maintaining the ad machine

New comment by whiplash451 in "Notes from the Mistral AI Now Summit"

whiplash451 — Sat, 30 May 2026 04:42:39 +0000

Maybe because distilling small models from bigger ones that you control gives you better small models than fine-tuning from bigger models you don't control?

(I am not claiming it is the case, but stating this as an assumption)

New comment by whiplash451 in "Can we have the day off?"

whiplash451 — Thu, 28 May 2026 04:43:45 +0000

Companies don’t care about it but governments certainly do.

Over the past 150 years, the work week dropped from 70+ to 35h in France.

Granted that’s a long time horizon but still.

New comment by whiplash451 in "Dropbox CEO Drew Houston to step down"

whiplash451 — Tue, 26 May 2026 21:45:27 +0000

The point is that we should be directing our energy in a direction that’s net useful for human kind which should translate into growth. Dropbox is not one of those because there are many viable alternatives.

New comment by whiplash451 in "The real cost of owning a home"

whiplash451 — Tue, 26 May 2026 21:21:06 +0000

You are also reminded of the same dichotomy when you (the renter) get a letter from your landlord asking you to leave the premises you’ve liked a lot for the past few years.

New comment by whiplash451 in "A successful Japanese trial of a ramjet engine designed for Mach‑5 aircraft"

whiplash451 — Mon, 25 May 2026 20:36:54 +0000

This is already a solved problem for the class of customers they are going after.

New comment by whiplash451 in "The Art of Money Getting"

whiplash451 — Sat, 23 May 2026 16:02:33 +0000

> low in extraversion and agreeableness

I don’t know that these are awesome features for an engineer. There’s a big unsaid cost to this in my experience

New comment by whiplash451 in "Microsoft starts canceling Claude Code licenses"

whiplash451 — Sat, 23 May 2026 14:29:40 +0000

Opus 4.7 went through a major degradation a few weeks ago (way more hallucinations and rabbit holes than usual). Anthropic fixed it. Give it another shot.

New comment by whiplash451 in "I’ve joined Anthropic"

whiplash451 — Tue, 19 May 2026 20:04:33 +0000

You can’t build “working with amazing people”. At least not in a short amount of time. I bet that this was a significant part of the decision for Andrej.

New comment by whiplash451 in "Claude for Small Business"

whiplash451 — Thu, 14 May 2026 09:58:50 +0000

> a UI that makes claude code accessible

Isn’t that literally Claude’s web UI?

New comment by whiplash451 in "Arena AI Model ELO History"

whiplash451 — Thu, 14 May 2026 08:41:30 +0000

Neat. Would you add the option to normalize the elo over time (e.g update the model used as an anchor for the elo computation) so the diff between labs is more visible?