Thadd Selden

Maker, sailor, appreciator of the natural world. I live in Sacramento, CA with my wife, daughter, and house full of pets

Pronouns: he/him
Website: https://thaddeus.net
Github: https://github.com/thadd
Mantra: Move slow and fix things

Last Notes

2025-12-01 21:15:32 UTC

Anti-vax is a eugenicist ideology.

2025-10-04 22:16:43 UTC

- reply

Ultimately your AI proselytizing comes across almost the same way as the crypto boosters did.
This life-changing technology was started at OpenAI to save the world. Their latest-greatest project is a social media deep fake generator that burns ungodly amounts of power.
No one is using AI to save the world, they're using it to build their fortunes so when the world ends they can hide in their bunkers.

2025-10-04 22:12:13 UTC

- reply

And I'd say you either need to talk to more people *not* working in the tech industry or you've read too much tech reporting because: enterprise adoption is abysmal (because the ROI isn't there), there are almost no actual documented studies showing AI improves productivity (because the ROI isn't there for *most* cases, yours is the exception), and while the average Joe has played with AI, surveys show that very few use and rely on it the way you describe "almost everyone I talk to" doing.

2025-10-04 22:07:53 UTC

- reply

The future of AI is a massive collapse of the data center market with an eventual slower adoption of edge AI running on local devices. DC AI will remain for niche use cases but it will not be as you know it now.
AI will be what robotics are now: Personal use cases analogous to vacuums and drones with enterprise use cases analogous to factory automation.
But the massive cloud-based reasoning models at affordable prices will be as inaccessible as advanced bipedal personal use robots.

2025-10-04 22:02:34 UTC

- reply

I have used these tools and I have seen the utility in them. Gemini is used throughout my org.
But I also know what it takes to actually run them (I work for a company that sells software infrastructure that is used to run AI at enterprise scale).
And I've talked to numerous people and customers and know that the use case you describe is not one that has an ROI even remotely close to what it costs to run, by orders of magnitude.

2025-10-03 13:33:23 UTC

- reply

How depressing if we end up in a world where we prioritize a brainstorming tool over fixing climate change. Hell, we’ll probably drill the planet dry to power the GPUs because Grok told us global warming is a hoax.

2025-10-02 18:18:02 UTC

- reply

That's what all the optimists say and hope, and yet look at energy bills right now. They're skyrocketing. A huge part of that is energy diverting to data centers.
Yes we have capacity, for now, but with gigawatts more data centers (assuming they actually get built for customers that just aren't there right now), who knows?
Trump's return to coal may be welcomed by those being buried under power bills they can't pay.

2025-10-02 14:04:24 UTC

- reply

Read Ed Zitron’s work on the financing of the AI industry. Houses weren’t junk but the financing was. AI will go the same way and maybe soon.
AI investment literally needs more money than there is available in all of private equity and venture capital. Nvidia is selling GPUs to companies that then leverage those GPUs into debt to buy more GPUs, backed by Nvidia.
It’s a house of cards that will inevitably collapse. The question is what’s left on the other side and who picks up the bill.

2025-10-02 00:49:10 UTC

- reply

My sole point is that you’re very harsh towards AI detractors because you would pay a lot for the capability. But businesses won’t. Individuals are simply not enough to sustain the entire tech sector that has gone all in and taken the stock market with them.
We don’t know what it will cost once the bubble pops so you may be able to or you may not.
The subprime crisis didn’t lead to long-term affordable housing and the finances of AI bear much more resemblance to it than the dotcom era.

2025-10-01 17:06:02 UTC

- reply

There’s two issues this leads to:
1. The use cases you value it for really don’t have enough ROI behind them for the cost of building and running the models, even with economies of scale. I often make the analogy of highly advanced robots being very useful but economically inaccessible to most. AI is effectively subsidized in an unsustainable way right now.
2. Gen AI has to be hyped as a panacea to try to mitigate the first point, so it’s shoehorned it into many ill-suited use cases.

2024-12-23 20:39:52 UTC

Finished the clock in time for the gift exchange.
#maker #diy #lasercutting #woodworking
https://selden.social/system/media_attachments/files/113/704/067/926/832/811/original/050fc0588e7fa6d3.jpeg