Hacker News: ttul

New comment by ttul in "1-Bit Bonsai Image 4B Image Generation for Local Devices"

ttul — Sun, 31 May 2026 21:28:26 +0000

Within a day, someone will have trained a LoRA for this 1-bit model that enables hentai content generation on your Apple Watch.

New comment by ttul in "DeepSeek reasonix, DeepSeek native coding agent with high caching and low cost"

ttul — Sun, 24 May 2026 18:47:17 +0000

What you’re experiencing is the difference in model intelligence. Most models can seem pretty good at simple stuff over short time horizons. Complex work requires that more intelligence be stuffed into those trillion-dimensional spaces.

New comment by ttul in "Show HN: Agent.email – sign up via curl, claim with a human OTP"

ttul — Fri, 22 May 2026 05:46:00 +0000

Abuse handling prowess will be the USP that defines the first $100M company in the “agent mail” space.

New comment by ttul in "CISA Admin Leaked AWS GovCloud Keys on GitHub"

ttul — Tue, 19 May 2026 14:53:01 +0000

Yet another argument for the death of the API key. Replacements abound; let's get on with it.

New comment by ttul in "OpenBSD 7.9"

ttul — Tue, 19 May 2026 14:08:47 +0000

Developed at 4500ft elevation in the Texas of Canada, primarily.

New comment by ttul in "Free Energy from the Vacuum? NASA Pioneer Unveils Battery-Free CPU"

ttul — Fri, 15 May 2026 17:23:43 +0000

Someone smarter than me can probably answer this hypothetical: If scaled up, would this device become a heat pump? Extracting energy from the electromagnetic field via the Casimir effect would seem to transform random field fluctuations (i.e. "heat") in the environment into the movement of electrons down a conductor. If those electrons are sent away to do work elsewhere, to my mind that's a transfer of heat from one place to another. Moving 0.5W of heat isn't very interesting, but I wonder if there's a scaling up that would let you move meaningful amounts of heat around in this manner while generating electricity as a "bonus".

The market for moving heat around is pretty enormous.

New comment by ttul in "Haiku"

ttul — Wed, 13 May 2026 17:09:52 +0000

It's a shame that Be failed. I think they were a victim of Microsoft's aggressive anti-competitive activities in the late-1990s, combined with Apple deciding to bring back Steve Jobs via the acquisition of NeXT (making Apple a serious competitor in the same segment that Be was targeting -- multimedia and realtime applications). Ultimately, they prevailed in winning about $24M from Microsoft, but that was after the company had shut down. I presume the winnings went to Palm. Super cool to see Haiku continuing to develop. No doubt agentic coding is making it far easy for enthusiasts to improve and maintain projects like Haiku and I look forward to seeing where this project goes. You never know...

New comment by ttul in "Cloudflare to cut about 20% of its workforce"

ttul — Fri, 08 May 2026 03:55:05 +0000

I know this is cold comfort, but in times like this, it can be a good idea to start your own company. Cloudflare itself was founded in the wake of the GFC (post-2008), when tech was dead as a doornail. The best time to start something is when awesome people to work with are unemployed.

New comment by ttul in "Cloudflare to cut about 20% of its workforce"

ttul — Fri, 08 May 2026 03:49:59 +0000

They have a reputation to maintain, otherwise it will be difficult to recruit the best people in future. That being said, damn, that is a very generous package by any measure.

New comment by ttul in "Canvas online again as ShinyHunters threatens to leak schools’ data"

ttul — Fri, 08 May 2026 03:46:13 +0000

But, then, how would Trump’s family and cronies get paid?

New comment by ttul in "Agents can now create Cloudflare accounts, buy domains, and deploy"

ttul — Thu, 07 May 2026 00:17:30 +0000

Technically, you could create many top level Amazon accounts, but if you want to send lots of mail, you must warm up your account. So accounts can be created, but it’s useless if you need to send high volumes of messages.

New comment by ttul in "DeepSeek V4 Pro at 75% off until 31 May"

ttul — Thu, 07 May 2026 00:16:01 +0000

* from Chinese labs

New comment by ttul in "Agents can now create Cloudflare accounts, buy domains, and deploy"

ttul — Wed, 06 May 2026 12:39:20 +0000

I'm convinced that one of the top use cases for OpenClaw is orchestrating cold outreach email campaigns, as if there's nothing wrong with using AI to spam people to death. Platforms that enable sending cold emails are taking a sizeable risk that the low engagement of such emails stimulates some worsening inbox deliverability for the rest of their traffic (see [1] - you can't hide just by sending through big tenanted platforms like Amazon).

[1] Every message sent from Amazon SES carries a "Feedback-Id" header that allows Google (and anyone else) to track the Amazon account responsible for the message. The fourth field is an opaque but stable identifier associated with your Amazon account; receivers can and do use this for rate limiting: https://aws.amazon.com/blogs/messaging-and-targeting/underst...

New comment by ttul in "AI Product Graveyard"

ttul — Tue, 05 May 2026 14:08:00 +0000

Yes, I just booked a meeting using Reclaim. And, is it really “AI”? It’s a rules-based scheduling app.

New comment by ttul in "Claude.ai and API unavailable [fixed]"

ttul — Thu, 30 Apr 2026 01:39:49 +0000

I've been a Codex devotee since around last August. I don't know why everyone is so bonkers about Claude Code. It's not the only belle at the ball. Codex is rock solid.

New comment by ttul in "Waymo in Portland"

ttul — Tue, 28 Apr 2026 19:27:19 +0000

If the back window can be smashed in, that would also fit Portland aesthetics.

New comment by ttul in "Microsoft and OpenAI end their exclusive and revenue-sharing deal"

ttul — Mon, 27 Apr 2026 21:45:31 +0000

If you do the math (I did), in 2 years, open source models that you can run on a future MacBook Pro will be as capable as the frontier cloud models are today. Memory bandwidth is growing rapidly, as is the die area dedicated to the neural cores. And all the while, we have the silicon getting more power efficient and increasingly dense (as it always does). These hardware improvements are coming along as the open source models improve through research advancements. And while the cloud models will always be better (because they can make use of as much power as they want to - up in the cloud), what matters to most of us is whether a model can do a meaningful share of knowledge work for us. At the same time, energy consumption to run cloud infrastructure is out-pacing the creation of new energy supply, which is a problem not easily solved. I believe scarcity of energy will increasingly drive frontier labs toward power efficiency, which necessarily implies that the Pareto frontier of performance between cloud and local execution will narrow.

New comment by ttul in "Google plans to invest up to $40B in Anthropic"

ttul — Sun, 26 Apr 2026 01:25:47 +0000

Yeah, it did just come out. I should have probably said "GPT-Pro" and left out the version :)

New comment by ttul in "Google Flow Music"

ttul — Sat, 25 Apr 2026 01:11:58 +0000

Here's my personal take on what I'll call the new realm of "AI art". Whether it's prompting a music model or an image model, there is a huge space for creative output, limited only by the human imagination. Sure, tossing in a single prompt and letting the model crap out something will produce "slop". But if you pour your heart into exploring the high-dimensional landscape of the model, you can find truly amazing stuff. This is no different than exploring the creative landscape of music, photography, and other forms of art in the pre-LLM era.

I find that people who rush to negative judgement of LLM-generated art are not going far enough in the creative process to properly judge just how much juice there is to be squeezed out of those 50-billion-dimensional spaces.

New comment by ttul in "Google plans to invest up to $40B in Anthropic"

ttul — Sat, 25 Apr 2026 00:41:04 +0000

This sounds like my office, but we're a bit more tilted toward Codex. I personally use Claude Cowork for drudge-admin work, GPT 5.5-Pro for several big research tasks daily, and the LLMs munge on each other's slop all day as I try my best to wrap my head around what has been produced and get it into our document repository -- all the while being conscious that the enormous volume of stuff I'm producing is a bit overwhelming for everyone.

We are definitely reaching the point where you need an LLM to deal with the onslaught of LLM-generated content, even if the humans are being judicious about editing everything. We're all just cranking on an inhumanly massive amount of output and it's frankly scary.