meerita

I wonder if having a good account plan, a good ruleset, and a nice graph of the entire platform would give me tremendous leverage with companies like that. Nothing beats a swarm of agents checking everything, running multiple probes, and finding the best solution for every problem, including generating all the questions stakeholders need to answer before proceeding.

I believe I would end up enjoying life more while everything gets solved through agentic development.

If I compare this with my previous companies (publicly valued companies), I would definitely be on vacation while still delivering perfect PRs and features in a matter of hours, instead of the 2-3 week timelines it took me and my team in the past.

If you build a nice writing guideline, you can impersonate Mark Twain if you want with any model.

Grammarly was a great idea in the pre-AI era. Now I bet they're losing a lot with so many AI solutions that are better integrated across every ecosystem.

Some people will argue that Grammarly still has a better suggestion system, but models have become increasingly good at writing. I think this product and its revenue model will become obsolete pretty soon.

I tested it with Claude Code, and I can confirm it's way cheaper, better, faster and less verbose than Opus 5.

The current tests confirms Opus 5.5 is not verbose. In fact, it outputs using ASD-STEM100 as I recommend. It's a blessing.

Compared to Opus 5, and others, I also found DeepSeek 4.1 Max to be really good and cheap. I am testing right now with Opus 5.5 and I feel it way cheaper than Opus 5!

My father built several projects in FoxPro. I was too young in the ’90s to remember much of it, but I’m sure he’ll be super happy to check this out. The kicker is that we’ll probably need to buy a floppy disk drive and dust off some old boxes to find them.

OpenAI, Antrophic and others are operating with 80% margins. They can lower the prices for a long while.

As long as it's not as verbose as Opus 5, I am quite happy with a better version that's also less expensive. I will test it tonight. Grok 4.7 was horrible, and for mundane tasks I am relying on DeepSeek Flash 4.1 with great success using OpenCode.

It seems that these things goes on depending on each type of project. I've been reading reports and Grok 4.7 sucks at 3D, for example, really bad.

Define easy. I am running 40+ probes to crack the best compression algorithm, and I spent $3. If you surgically tackle the problems, you can do really complicated work for less money than by assuming a frontier top model will one-shot everything.

Well, at least I spent lots of dollars, and I had to use those models the same way I am using local and cheap models, with the same results.

By no one. For the price of 1M token you can get more and with better results with other models.

Grok it's really expensive. I'm getting really amazing results using DeepSeek 4.1 Flash for fraction of the price.

A good explanation of what is happening. It's wild. Peak APM was 6k. You can see the game.

- https://www.youtube.com/watch?v=1vsTqNwHquE

832K + 469k of tests.

True! I edited.

I wrote to the post of Andrea (a dev from the Copilot team) about their 800K LoC: 128 PRs, shipped incrementally. Existing end-to-end tests ran against the new code at every step.

Total: ~1,301,378 lines of Rust.

Production: 832,378 Unit tests: ~469,000 Combined: ~1.30 million lines

On top of the 832K LoC are mostly tests she answered:

> Yeap, 832,378 lines of production Rust. the +800K number is production only; unit tests are another +469K on top.

https://x.com/acolombiadev/status/2100660224298193081?s=20

Sorry, I just see a physical keyboard on a phone, and I cannot be convinced that it's a good idea. We're past that. If it were truly minimal, it wouldn't have a physical keyboard. It would have a full-screen display with a minimal but useful OS and UI/UX.

There will be a point where companies will not need to scrape any content. Agents will create endless streams of probes, and they will end up solving all kinds of knowledge problems.

Oh yes!

I never imagined how to pronounce it until now, "commander taco", isn't? I remember the old days of Slashdot all pools included CmdrTaco in one option.

(I felt the pain of TeX so much). There's an enormous opportunity to rewrite all kinds of old projects in Rust. I wrote my own lsof implementation, and the benchmark results are incredibly good. I love this new era where you can take something that works but was built 30 years ago and make it 50x faster, better, and more efficient.

I know this pain because, right now, I have zero PRs opens, yet the PR tab is always saying I have 1. It's plainly stupid.

Incredibly expensive! They must be using frontier models to perform really dumb tasks.

Am I right of the token expending is +4K?

I am working on several projects:

- Puma, The terminal browser: https://github.com/meerita/puma-browser

- Kernq, a full replacement of lsof: https://github.com/meerita/kernq

- Entroq, modern lossless compression for modern systems: https://github.com/meerita/entroq

Why not? The story seems so far more interesting!

These CSS layout techniques remind me of others in the past where you put a lot of effort and barely 2% of your userbase will use it.

China will not stop and will take over the rest of the world.

That is not quite the same problem. Modern browsers are vastly more standards-compliant than they were in the IE era, and most CSS/JS works across Chrome, Firefox and Safari without browser-specific hacks.

There was a time when we routinely did browser sniffing, maintained IE-specific code paths, used vendor prefixes everywhere, and sometimes shipped completely different fixes depending on the browser. That largely disappeared as browsers converged on standards.

Chrome-first development can still cause bugs elsewhere, but that's usually a "testing/implementation problem" rather than the same compatibility mess we had back then.