shmolyneaux

Whether you're pro-AI or anti-AI, everyone benefits from non-AI-written content. The anti-AI crowd benefits from avoiding content they dislike and the pro-AI crowd gets more quality training data.

I would way rather read handwritten articles. I think everyone should celebrate the human craft of writing. I really hope this doesn't slowly become the "no-VFX" posturing from the film industry, however.

I love plain text, but the calculus of being able to access the content in 50 years is much more interesting in the age of agents. They can infer the meaning of structured data without schemas, decompressed archives, find embedded files, etc. It's not perfect, but the durability of binary formats is better now than it has ever been.

It's not baked-in to the weights of any model, but agents can write their own tools to work with arbitrary binary formats and get many of the benefits of off-the-shelf unix utilities.

I love that the Godot game engine has a textual scene description. That's a stark contrast to Unreal Engine's binary format for blueprints (visual scripting).

I definitely appreciated your post about Nano Banana Pro. It's also a genuinely useful time-capsule for how these systems evolve and where they fall short. I've preferred the output of ChatGPT Image 2. I think a post would be very helpful for folks to see what they're missing.

There is a bit more to the story than a 1% wealth being "equivalent" to a 20% income tax. The primary difference is that unrealized gains are taxed by a wealth tax. We need a mechanism for assets to be sold by the richest in society. If those with assets keep accruing more assets the median person will suffer. When we're talking about real assets (housing, retail shops, warehouses, land) we don't need to be concerned about capital flight. The assets are still there on the ground. Reducing the cost of those assets is exactly what we need to help a local economy.

That being said, the richest are effectively _not_ paying the highest marginal tax rate considering all the tax structuring they do. Claiming that they would be paying the highest income tax in the world is misleading, for one. Secondly, the richest in the world _should be_ paying the highest income tax.

I think more people need to see this. This is how the creator of C++ thinks we should be writing code. This is what he thinks code should look like. To split a string by whitespace we should use `while (cin >> s)`. We should have a `typedef` in the middle of functions. Iterations should use `.begin()` and `.end()` everywhere. There might even be a bug with a trailing "+" appearing in the output?

Imagine if this was a new language that the dev community was seeing for the first time. It's hard to imagine it gaining much traction.

It's true that it's shorter, but I suspect that the if-return, if-return pattern compiles down to much faster code. Separately, this code was originally written in C then ported. There are reasonable explanations for why Matz has the code written this way besides the typical AI slop.

Github Copilot changed to token pricing earlier this week

Opus was the reason I paid for GitHub copilot, but they had the pricing model completely wrong. I could assign copilot to a substantial issue using Opus and have it handle 30 minutes of work with many subagents with iterative testing. With 300 "premium requests" a month I could have copilot do substantial work for 3 premium requests per issue. It was very clear that this was unsustainable for Microsoft to pay for, so I expected change to come.

However, I never expected Opus 4.6 to be removed for the cheaper plans. I expected the pricing model to change, but not to lose access to the model. Moving to being token-based makes sense. It makes the cost more closely aligned with user pricing.

It was nice while it lasted. I got Opus 4.5 to do a lot of work from the beach by assigning it to detailed issues. With this news I've cancelled my Pro subscription. That will help a bit with their capacity issues.

That seems like a GDPR violation waiting to happen. It shouldn't be possible for them to store an email address like that forever and be in compliance.

That's noted further down the page:

- `_*` = any string

This was a fun read, but it's a shame it hasn't gotten more attention. I wouldn't have guessed that Bun was using a custom fork of Zig. Since the feature has been closed, I strongly suspect that they'll eventually need to need to go back to upstream and revert their usage of private members. Using Python style `_members` seems like it should be good enough. If you're already manually managing memory it seems like respecting those privacy hints should be reasonable enough.