Hacker News: rryan

New comment by rryan in "Gemini 3.8 text-to-speech"

rryan — Thu, 24 Sep 2026 01:12:45 +0000

HN is not your therapist, but possibly your decision making is driven by branding instead of judging technology on its merits? Easy mistake to make ;)

New comment by rryan in "Teaching my kid to code with a modern MUD"

rryan — Tue, 18 Aug 2026 15:51:02 +0000

CircleMUD was my first intro to C and network programming. I was at a computer camp for 8-10 year olds that taught BASIC. One of the counsellors was doing their undergrad at caltech and pulled the students who were bored by BASIC into a separate group to learn C by editing CircleMUD.

Hacking up a pre-existing MUD was the perfect mix of immediacy / interactivity and fun to hook me. So thankful to that guy, wherever he is. He probably changed the course of my life.

New comment by rryan in "Retire the Abstractions"

rryan — Wed, 12 Aug 2026 04:19:34 +0000

Heh, not really

New comment by rryan in "Retire the Abstractions"

rryan — Wed, 12 Aug 2026 04:16:14 +0000

Disagree. The abstractions they are talking about aren't like, the y-combinator. They're talking about boring software engineering abstractions.

New comment by rryan in "Retire the Abstractions"

rryan — Wed, 12 Aug 2026 01:57:19 +0000

I know nothing about the PhD candidate and professor who co-authored this, but I don't think this statement makes sense in general.

Typically CS labs at universities contain people who have not spent a significant amount of time exposed to large industry codebases and the corresponding complexity. I think the post would have more credibility coming from e.g. the platforms team at a tech company with a monorepo.

New comment by rryan in "Retire the Abstractions"

rryan — Wed, 12 Aug 2026 01:49:04 +0000

agents are unlike humans in many ways but they have their own limitations. it's magical thinking to believe that they are able to cope with zero software abstractions, especially as a codebase increases in size and complexity

and today's agents increase codebase size and complexity much faster than humans do via traditional software engineering methods, so this is likely to hit a wall faster than if human software teams were to do this experiment.

New comment by rryan in "Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs"

rryan — Wed, 05 Aug 2026 21:21:50 +0000

No longer the case

New comment by rryan in "GPT‑Live"

rryan — Wed, 08 Jul 2026 18:00:33 +0000

ehh, no? definitely not.

https://kyutai.org/blog/2024-07-03-meet-moshi/

New comment by rryan in "Moving fast in hardware: lessons from lab to $100M ARR"

rryan — Tue, 07 Apr 2026 19:45:41 +0000

As a human emdash enjoyer, I'm pretty damn grumpy that everyone assumes I'm a chatbot now. :P

New comment by rryan in "Moving fast in hardware: lessons from lab to $100M ARR"

rryan — Tue, 07 Apr 2026 16:51:09 +0000

I think the super long timelines in automotive make it really hard to succeed. Really impressed with what ClearMotion accomplished given that

Moving fast in hardware: lessons from lab to $100M ARR

rryan — Tue, 07 Apr 2026 15:07:38 +0000

Article URL: https://blog.zacka.io/p/simplify-then-add-lightness-bc4

Comments URL: https://news.ycombinator.com/item?id=47676557

Points: 116

# Comments: 26

New comment by rryan in "The bitter lesson is coming for tokenization"

rryan — Wed, 25 Jun 2025 05:38:50 +0000

Don't make me tap the sign: There is no such thing as "bytes". There are only encodings. UTF-8 is the encoding most people are using when they talk about modeling "raw bytes" of text. UTF-8 is just a shitty (biased) human-designed tokenizer of the unicode codepoints.

New comment by rryan in "Transformers Without Normalization"

rryan — Sat, 15 Mar 2025 06:50:46 +0000

RMSNorm is pretty insigificant in terms of the overall compute in a transformer though -- usually the reduction work can be fused with earlier or later operations.

New comment by rryan in "Can Gemini 1.5 read all the Harry Potter books at once?"

rryan — Sat, 13 Apr 2024 04:30:32 +0000

ML 101: Do not evaluate on the training data.

Yes of course it can, because they fit in the context window. But this is an awful test of the model's capabilities because it was certainly trained on these books and websites talking about the books and the HP universe.

New comment by rryan in "(next Rich)"

rryan — Fri, 04 Aug 2023 15:03:08 +0000

Thanks for everything, Rich. You inspired me repeatedly.

New comment by rryan in "Foundation models are going multimodal"

rryan — Thu, 27 Jul 2023 20:43:12 +0000

https://elevenlabs.io/

https://twelvelabs.io/

.... wat

New comment by rryan in "Google doesn’t want employees working remotely anymore"

rryan — Thu, 08 Jun 2023 05:07:25 +0000

Yea uh, don't know what was going on there but there are roles with direct reports that aren't in mountain view (unless you mean, like in 2004). NYC has 1000s of googlers.

New comment by rryan in "AudioGPT: Understanding and Generating Speech, Music, Sound, and Talking Head"

rryan — Wed, 26 Apr 2023 06:32:09 +0000

This is ... not what I expected. It's basically wiring up pre-trained models to ChatGPT via a router and "modality transformations" (a.k.a speech-to-text and text-to-speech).

I expected it to be a GPT-style model that processes audio directly to perform a ton of speech and maybe speech-text tasks in a zero-shot manner.

New comment by rryan in "$6.96B was raised by private longevity companies in 2022"

rryan — Sat, 28 Jan 2023 03:40:40 +0000

Except when your doctor won't prescribe it you offlabel because you don't have a condition they were taught about in med school

New comment by rryan in "Ask HN: Google spam filter getting worse?"

rryan — Fri, 20 Jan 2023 17:41:33 +0000

In the past few months I've been getting much more fake order / payment spam.