toddmorey

It's a fascinating case of use our product to build websites & apps until we can get our product to entirely replace websites & apps.

AI is this strange double-threat for folks in software development: attacking both the demand and supply at the same time.

Me, I don't think they are super eager to be a hosting platform even though this is a very logical extension for them. I think they're hoping this is a catalyst to make login with OpenAI as ubiquitous as login with Google.

Except for critical services like taxes & benefits

That seems like the intent for sure… but behind the scenes, they are really the LLM with a hosted sandbox. I’m still thinking any sort of moat will be difficult & agent services that can use multiple LLMs will be popular.

I had this naive assumption for years that MacOS was the most stable / workable *nix desktop available, but I’d been sleeping on modern Linux distros.

Been on Mac since the 90s, now a very happy Linux desktop user.

I kinda think home charging is important. You never know when the public charger will have no spots.

Aside: Quanta magazine always has excellent design & visuals.

Author here. This started as a 20-second ad: a household robot waters a plant, the camera orbits, and its revealed the water is actually missing the plant. Fun gag, but I had no idea the struggle that would ensue trying to get the model to produce a result against its rewarded training data. Hadn't done much with generative video before this; Learned a lot about the state of video models and AI world knowledge on this project.

I mean, it's a structured output model that (apparently) can't hallucinate. I don't mind the name.

From the demo video: "Welcome to the team, we're looking forward to working with you" is sooo creepy in a synthetic AI voice. In general, try not to have agents express sentiment that really should come from a human in your company.

How long until "conversations are not shared with Meta's ad systems" quietly disappears from the site?

I almost think they would have done better to launch this as more disconnected sub-brand. I think Meta / Facebook are actual brand baggage for this sort of thing, even outside of tech circles.

I do think models currently don't have enough source skepticism.

If you look at agent traces when asked to compare two options to help inform a decision, many of the comparison pages cited in research are often hosted by one of the companies being compared; nearly all are AI-generated AEO plays. Not deeply considering the motive of published information is currently a glitch that can be exploited, but the window will close.

I'm sure model providers will set up some crappy pay for play verification system for "trusted" product information, comparisons, and reviews.

Luckily he does not enjoy anywhere close to 50% support

Yes, this is a horrible practice. Not just Amazon… most major e-commerce platforms set up fake pages for products they don’t offer to try to get you to the products they actually do.

Google needs to ban this practice because it makes their results garbage.

Have always joked with colleagues about how hard it is to find content in Google Docs, the office suite built by a search company.

Yeah the difference in token usage across difference models they found in the article was broader than I expected, but I was sort of floored by the variance in token usage for the same prompt (run multiple times) with the same model.

As someone who's done web development for 20+ years, I find the model personalities pretty dang fascinating, especially how they develop (and evolve) design sensibilities.

I'm interested to know how the classic AI "purple preference" emerged (organically?) and if the beige wave came out of specific training efforts to combat it?

To your point on development work (the code itself), I was talking to some friends on the Google Chrome team about any research understanding the model's preferences around framework ergonomics and abilities to properly implement core web standards for given tasks. I think that would be super fascinating.

You can’t do inter-city Waymo trips. These are different animals.

Agreed, but I think people can imagine and companies are very motivated: powertools, phones, laptops, watches, backup UPS, cars, hospital equipment... there's high demand for durable, long-lasting batteries. I think the research is there, it's just complex chemistry. We'll get there. Impressive to see the progress in EV batteries and they've actually turned out to be more durable than first feared.

Have we ever as a species successfully changed our behavior on a scale needed to address something like climate change and underwater oxygen loss?

What would be the closest “we’ve got to change or perish” moment we’ve successfully navigated?

Nope, HTMX could power endless scroll features if you want them

One way to think about it is that more powerful models mean solid best practices are more important than ever, so humans moving too quickly / carelessly bites us more than ever.

If a typical SaaS platform moved and shifted this fast with this many downstream consequences, we’d tell them to slow the fuck down, stop launching new features, and focus on security for a second.

But in AI I guess the idea is that more power and intelligence will solve for everything else.

"We believe in a world where we can have safety and privacy."

  - Flock camera feeds were set up public-facing, giving anyone on the internet access to live high-resolution feeds
  - Flock's AI does more than just read license plates. It catalogs vehicle color, make, model, bumper stickers, dents, and roof racks.
  - Police officers nationwide have abused their access to the Flock network to inappropriately track romantic partners, neighbors, and personal acquaintances
  - local police departments share Flock-captured license plate data with federal immigration agencies like ICE and CBP
  - Flock cameras have been used to track lawful citizens attending political demonstrations
  - Flock also expanded into audio surveillance (!!) before public outrage made them back down a bit

Yeah I think the podcasts were just a clever hack until AI is fast enough to support realtime. It’s practically there now, so a fake podcast is a funny relic and hard to sit through.

I can’t believe how bad Claude’s voice mode is. So much latency and it often tells me “this is rather technical” and would be better talking it out over text.

OpenAI has much less latency, can adapt to speak about any topic (though will tell you it put a code sample to look at later), and it just feels like brainstorming with a colleague.

If you haven’t tried the live voice version of ChatGPT recently, you should. I love to brainstorm ideas while on walks and it’s fabulous for that.

Ask it to summarize the conversation into a doc that will be waiting for you when you are back at your desk.

I did too! That HAD to be by design.

This is awesome! I made a Soroban (Japanese abacus) version, but I couldn't add it to the collection since I used css-based animations. Published here: https://soroban-clock.netlify.app

I'm very pro tech. Because I'm pro tech, I honestly wish there were more ethical companies in the space. It can be hard to find US tech companies to cheer for. My main business-related challenge with Elon is his public predictions are so wildly ungrounded. Skepticism is absolutely warranted.

Two things:

  - You can cheer for his companies and technologies but against these financial maneuverings because they risk genuine harm & setback to the industry

  - I stopped asking for people to put politics aside in support of these companies when I realized Elon couldn't put politics aside to support these companies.

But it's not about whether it was AI or human authored. That misses the point. We're all fatigued on the writing style. The same cadence and patterns; the same phrases and terms like "load-bearing". Used everywhere, they create a super fatiguing monoculture in all the writing. It's like if every illustration on the internet suddenly contained Garfield the cat.

I thought I read somewhere that according to filings for going public, subscription revenue is tiny… like 5%.

Edit: consumer Claude subs are the 5%. I’d bet most all of CC subs lump in under enterprise.

  - API & Enterprise: 75% to 85% of total revenue.
  - Business Subscriptions: Roughly 10% to 15%.
  - Individual Subscriptions: About 5%.

Processing markdown into html is a significant task for many blogs, sites, and apps. Satteri moves that processing into rust for speed gains.

Now used in Astro 7: https://astro.build/blog/astro-7/#markdown--mdx-in-rust