Hacker News: thaw13579

New comment by thaw13579 in "There are a few things that I look back on as my mistakes in the early days"

thaw13579 — Wed, 24 Jun 2026 17:47:11 +0000

That's literally treating people as means to an end, when it's not clear if the "end" even required sacrificing young people's lives. In the grand scheme of things, why not spend an extra year to build at a pace that doesn't burn people out? If you look at the careers of the people who subsequently left, by Sandy's account, they went on to to run their own great game studios, so it's not clear that a grind was necessary.

New comment by thaw13579 in "Midjourney Medical"

thaw13579 — Thu, 18 Jun 2026 16:55:54 +0000

They haven't shared enough to say anything concrete, but one important detail is that this system doesn't have many of the benefits usually found in ultrasound (portable, cheap, simple). From what they've shared, it's a large water bath that patients need to be submerged in with a large array of US sensors with high costs in data and reconstruction compute. It might have lower upfront cost, but otherwise, I personally don't see the advantages yet. A typical MRI can be done is under 30 minutes in street clothes, and personally I'd rather not deal with the logistics of a water bath...

New comment by thaw13579 in "GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers"

thaw13579 — Fri, 23 Jan 2026 10:30:04 +0000

To be honest, this one could just as well be a sloppy bibliography.

New comment by thaw13579 in "X blames users for Grok-generated CSAM; no fixes announced"

thaw13579 — Mon, 05 Jan 2026 20:13:04 +0000

The core issue is that X is now a tool for creating and virally distributing these images anonymously to a large audience, often targeting the specific individuals featured in the images. For example, to any post with a picture, any user can simply reply "@grok take off their clothes and make them do something degrading", and the response is then generated by X and posted in the same thread. That is an entirely different kind of tool from an open-weight model.

The LLM itself is more akin to a gun available in a store in the "gun is a tool" argument (reasonable arguments on both side in my opinion); however, this situation more like a gun manufacturer creating a program to mass distribute free pistols to a masked crowd, with predictable consequences. I'd say the person running that program was either negligent or intentionally promoting havoc to the point where it should be investigated and regulated.

New comment by thaw13579 in "My role as a founder-CTO: year 8"

thaw13579 — Wed, 31 Dec 2025 20:12:33 +0000

Those were most likely from a multi-line list that was converted to a single line when the comment was processed on submission.

New comment by thaw13579 in "Brown/MIT shooting suspect found dead, officials say"

thaw13579 — Sat, 20 Dec 2025 06:00:21 +0000

Sadly so, students often associate their self-worth with research and academic achievement, so if things go south, for whatever reason, they are in crisis.

New comment by thaw13579 in "Oliver Sacks put himself into his case studies – what was the cost?"

thaw13579 — Thu, 18 Dec 2025 22:19:12 +0000

Curious why this comment is being flagged if anyone minds explaining.

New comment by thaw13579 in "We Induced Smells With Ultrasound"

thaw13579 — Mon, 24 Nov 2025 04:39:51 +0000

I agree there are some red flags here to me. One is the priority claim "As far as we know, no one seems to have done this kind of stimulation before - even in animals." The other is the definitive conclusion based on weak experimental design and documentation, "Can ultrasound make you smell things that aren’t there? Turns out, yes!"

These are big scientific claims, but the work is clearly too premature to make those conclusions, and it lacks the connection to prior work and peer review needed for making priority claims. It's really great hacker-tinkering work though, and it could turn into solid science if they take more care with it.

If this effect is real and truly novel, my cynical expectation is that someone already established in focused ultrasound will read this, apply a more rigorous approach, and get the recognition that they are hoping for through more establish channels.

New comment by thaw13579 in "We Induced Smells With Ultrasound"

thaw13579 — Mon, 24 Nov 2025 04:27:55 +0000

I wonder where they got their equipment and research space. A charitable explanation is that they purchased it out of their own pockets, but otherwise, they really should acknowledge their support if it's from a university, federal grant, foundation award, etc. In my opinion as someone with domain experience, they don't show any novel solutions to accomplish this, it's mostly just that they have the time and resources to experiment try out, so it's especially important to acknowledge who enabled it.

New comment by thaw13579 in "Claude Is Down"

thaw13579 — Fri, 07 Nov 2025 16:26:25 +0000

Probably an M4 which has up to 128GB currently

New comment by thaw13579 in "Who benefits from the MAHA anti-science push?"

thaw13579 — Wed, 22 Oct 2025 17:59:38 +0000

An experiment is essentially a way to question ones own beliefs by probing how well they align with reality. There are some theoretical scientists, who don't experiment, but I think they also benefit from counterfactual reasoning to do their work.

New comment by thaw13579 in "Who benefits from the MAHA anti-science push?"

thaw13579 — Wed, 22 Oct 2025 17:55:31 +0000

Questioning is great, but to generate scientific knowledge, we need a few more steps, roughly speaking:

1. Ask a question 2. Form a hypothesis 3. Experiment to test it 4. Analyze results 5. Draw conclusions 6. Repeat

The MAHA folks essentially disregard this as a valid process for gathering knowledge. They occasionally talk about experiments and studies, but they are selectively chosen to support their conclusions in a posthoc way, ignoring both evidence to the contrary and basic methodological issues. When people describe them as "anti-science," I believe this is the kind of thing they have in mind.

New comment by thaw13579 in "Sequoia COO quit over Shaun Maguire's comments about Mamdani"

thaw13579 — Wed, 22 Oct 2025 17:39:16 +0000

That was previously the case from my perspective as well, but it seems to be rapidly changing.

New comment by thaw13579 in "Who benefits from the MAHA anti-science push?"

thaw13579 — Wed, 22 Oct 2025 17:35:48 +0000

It's especially hard given that big tech companies and their leaders are working closely with government and explicitly supporting certain political missions, there are few truly apolitical corners of tech now.

New comment by thaw13579 in "iPad Pro with M5 chip"

thaw13579 — Wed, 15 Oct 2025 18:29:53 +0000

Two reasons for are the cellular connection and smaller form factor. Would love to have a MBA with those

New comment by thaw13579 in "iPad Pro with M5 chip"

thaw13579 — Wed, 15 Oct 2025 18:28:00 +0000

Thanks for mentioning the non-secure backup in the UK, I missed the memo on that, and fyi, for any one else who did: https://support.apple.com/en-us/122234

New comment by thaw13579 in "What Americans die from vs. what the news reports on"

thaw13579 — Wed, 15 Oct 2025 16:13:32 +0000

I dug into it more and found the lab who conducted the study at Georgia Tech published their results in a reputable journal with peer review (https://www.nature.com/articles/s41599-021-00993-6). The analysis seems fairly evidence-based, consisting of "54 million social media posts, both pre- and post-drills in 114 schools spanning 33 states."

New comment by thaw13579 in "What Americans die from vs. what the news reports on"

thaw13579 — Tue, 14 Oct 2025 22:13:01 +0000

We of course should prepare and have protocols to protect children in these scenarios, but there are better and worse ways to go about it. I essentially believe it's okay to leave young children blissfully ignorant of low probability / high impact harms (there are many that are equally likely to school shootings that we ignore). Lockdown protocols and training seem fine to me, if they are sufficiently abstract, but there is an emerging trend of "crisis simulations" which involve people posing as shooters, simulating gunfire sounds, and staff / students posing as shooting victims, etc. I think adults can handle this kind of realism, but there is evidence for harm in young children.

https://www.nejm.org/doi/full/10.1056/NEJMp2301804

New comment by thaw13579 in "What Americans die from vs. what the news reports on"

thaw13579 — Tue, 14 Oct 2025 21:54:25 +0000

A summary from the Everytown report "The Impact of Active Shooter Drills in Schools"

"Active shooter drills in schools are associated with increases in depression (39%), stress and anxiety (42%), and physiological health problems (23%) overall, including children from as young as five years old up to high schoolers, their parents, and teachers. Concerns over death increased by 22 percent, with words like blood, pain, clinics, and pills becoming a consistent feature of social media posts in school communities in the 90 days after a school drill. "

https://everytownresearch.org/report/the-impact-of-active-sh...

New comment by thaw13579 in "What Americans die from vs. what the news reports on"

thaw13579 — Tue, 14 Oct 2025 21:10:03 +0000

It would be great to have a similar analysis for elementary school-aged children. Many schools are using "crisis simulation" of active shooter events in an effort to prepare for them (and presumably reduce the risk of death). While good natured, I think it's ultimately just needlessly traumatizing children, since school shootings account for <0.1% of deaths. While school shootings are devastating and sadly on the rise, the media greatly exaggerates the risks in people's minds. By the numbers, the biggest mortality risks for children are drowning and automobile injuries while unbuckled, both of which can be trained without inflicting psychological harm.