michaelbuckbee

Good catch, there was an issue with the second hardest thing in programming (caching).

Here's an updated eval with the proper models https://a3bmfqfom3.evvl.io/

My sister in law is a pharmacist and the heaviest non-dev ChatGPT user I know and her main use case is writing professionally polite messages to doctors on how the drugs they prescribed to a patient would have killed them had she not caught a particular interaction or common side effect.

There's a lot of "tone" in it as she's not trying to anger these folks, but also it's quite serious, but also there's just everything else happening in medicine.

Feels like a great use.

I did a quick eval comparing Grok 4.3, Opus 4.7 and GPT 4.1 and they actually seem pretty similar:

https://ofw640g9re.evvl.io/

They all did pretty well at a more "formal" tone, but GPT4.1 was the only one that didn't make me cringe with a "casual" tone.

[edit] fwiw, grok was also the fastest+cheapest model, claude was slowest and priciest.

This feels like a restating of the idea that for any given endeavor AI raises the floor of quality but doesn't push the ceiling.

Fwiw - I did a fairly large comparison of Gemini Nano (the in browser ai model) vs a comparable free hosted model of Gemma (from OpenRouter) and the hosted model absolutely trashed the local model on every aspect of speed, reliability, availability, etc. [1]

I'm not particularly happy about that outcome as I wish we had more locally run AI models for reasons of privacy and efficiency, so this is more just a warning that at present there are some severe tradeoffs.

1 - https://sendcheckit.com/blog/ai-powered-subject-line-alterna...

Maybe check out TRMNL, they've got a Home Assistant plugin.

From the article: "Polar bears are still sadly expected to go extinct this century, with two-thirds of the population gone by 2050,"

There's a push and pull here, Typescript + React + Vercel are also very amenable to LLM driven development due to a mix of the popularity of examples in the LLMs dataset, how cheap the deployment is and how quick the ecosystem is to get going.

So we have two things here:

1. "Mythical Man Month" which is the shorthand for a whole book + concept that you can't just throw more people at a software development project and get linear productivity improvements as the communications overhead (meetings, emails, mistakes due to poor assumptions, etc.) deeply eat into the raw number of productive hours that a new person added to the team brings.

2. AI automation tools (Claude Code) are often described as a "junior developer" which is an imperfect comparison as while you could potentially sort of set them up that way many people use them as more of a singular force multiplier.

I use them to work on many more projects in many more ways and ship far more than I could even if I had a "junior developer" sitting alongside of me as there's not the same level of communication needed.

Memory makers make capital investements (build different factories, convert physical production lines, etc.) to meet orders that have been place for the next ~5 years.

OpenAI (or whoever) crashes and can't pay for the order leaving the memory makers in a tough spot.

Partially this was solved by the ZipDrive being designed for portability as well (I'm not even sure if there was ever a built in model). So if you needed to copy a large file and take it to a friends house you just took the drive with you.

More importantly it collapses mythical-man-month communication overhead.

What's weird though is the bifurcation in pricing in the market: aka if your app can function on a non-frontier level AI you can use last years model at a fraction of the cost.

This is one of those slippery slope things where Grammarly did "just" Grammar and then slowly got into tone and perception and brand voice suggestions and now seems to more or less just want to shave everything down to be as bland as possible.

Is there an example of a consumer facing SaaS that's been able to handle the "unlimited" in a way you'd consider positive?

Two things:

1. Tests have always been both about the function of the application, but also the communication of what should be occurring to the larger team or yourself six months down the road.

With automated software development the communication with the LLM itself is a much larger part of it so I feel like it's "ok" to have lots of easy tests that are less about rigor and more about "yes this is how this should work"

2. Ideally we're going to get to the point where the tooling allows for adversarial agents with one writing code and one writing tests. Even for now just popping open a separate terminal window and generating+running tests in it from your main coding terminal is helpful.

I feel like AI just raises the floor but doesn't push the ceiling on the quality of creative works.

There's still space for creativity, novelty, invention and human intuition.

MCPs are for documentation. CLI->API is for interaction.

Very strongly reminds me of Mac Mini Colo - https://www.macminivault.com/mac-mini-colocation/ - it's a clever idea.

As doom and gloom as things are generally, I do think things have gotten better. Due to legislation and commercial pressure things like wifi routers shipping with the same default password and open settings have gotten better. Webhosts and ISPs have implemented many improvements to protecting their residential customers.

I take your point, but think that it's also maybe too far.

Bunny has a lot of that (they have a sqlite compatible API, edge functions, but the call it something different, etc.)

That being said, I had enough issues with Bunny and CF debugging across regions that I made this free tool to do both remote HTTP and TCP traceroutes to keep my sanity: https://dnsisbeautiful.com/global-http-availability

FWIW - I did a real world experiment pitting the built in Gemini Nano vs a free equivalent from OpenRouter (server call) and the free+server side was better in literally every performance metric.

That's not to say that the in browser isn't valuable for privacy+offline, just that the standard case currently is pretty rough.

https://sendcheckit.com/blog/ai-powered-subject-line-alterna...

I don't really understand the internal mechanics of of this, but my first thought was why not combine this with a linter/tests. So that it produces all the forks and only keeps the syntactically correct ones.

The underlying intent here (figure out if it's an existing customer of our locally installed apps when they visit our website) doesn't seem bad, but I certainly dislike both the hosts file and localhost detection options.

I'm curious if there's a "good" way to do this.

I build my own products and services and the effective ROI for paying for a more or less unlimited max Claude Code plan is fairly ridiculously positive.

Genuinely asking: has there been a case of the prediction markets being "right" or valuable about something outside of the norm?

Worth reading for the discussion of the LLM watermark technique alone.

Perhaps not so strange in that from a practical standpoint many "AI problems" are actually "Data Problems" when you start peeling things back.

That's a good trick (the "get ready" ping). It reminds me of how early Instagram was considered fast because they did the photo upload in the background while you were typing your caption so that by the time you hit "upload" it was already there and appeared instantly.

It was genuinely quite frustrating that all of the "normal" domain names were taken for this, but not actually used (basically everything "moniker.TLD").

What was surprising was that all the prefix+suffix variants of app/now, etc. were also taken so this was really just me trying to push it hard the other way.