protastus

The AI verification is not "handled structurally".

The vast majority of my tokens don't go into the authoring -- they go into review iterations. And the problem is that many reviews don't converge, and even the ones that do, can do it slowly, with drift and leave a residue.

Reviews don't converge when the premise is flawed. You can put tripwires to try to catch this, but it's not a robust process. Agents will generally try to act on a flawed premise, especially when they can't foresee the long-term consequences of the axioms under which they are operating.

Reviews drift because the author and reviewer influence each other to depart from the original problem statement. This is also correlated with unbounded scope creep and slop. Again, one can put tripwires to detect this and escalate but agents are extremely creative at finding new ways to ruin your day.

It appears so! https://www.vidaliaonions.com/

I bought a box of onions from him before and they were very good. I ate those onions for a long time.

The only way to stay sane with Claude is to use Fable. I let Fable spawn Opus review agents but I won't chat with Opus, and I won't let Opus write code.

I thought I could use Opus 4.6 but I gave up on it too. From a prose perspective it's much better than Opus 4.8 or 5. But it's still very verbose and Fable is dramatically more capable.

I'm sure folks from Anthropic love this conclusion because this setup is extremely expensive.

We doesn't need a signed affidavit on GitHub to trigger an investigation.

This already crossed the line of reasonable suspicion. The investigation is where evidence gets collected.

Who knows what other improper behavior these people have engaged in and what other secrets they have leaked, intentionally or by side effect.

When negligence is so bad that it looks like sabotage from a hostile agent, then criminal investigations are needed to learn more about the people who did it, the others who enabled it, and deter similar future acts.

DOGE did a lot of bad things, but it didn't force anyone to commit credentials to a repo, disable scanners to get away with it, and then make the repo public.

In 2026, storing government credentials in a repo and not having scanners to flag it should be investigated. I am highly suspicious of anyone doing this in a professional capacity. If I worked at a foreign intelligence agency and saw this, I would first think it's a honeypot, and an unimaginative one because it's so lacking in subtlety.

Speaking as a non-expert, I think Pix has much bigger scope. Pix is account-to-account. One can buy real estate, pay bills, make person-to-person and business-to-business transactions, government payments, recurring payments. The funds also settle instantly.

Most people don't experience the full scope of Pix which is impressive.

I think we need to put this in context for folks who are not from Brazil.

Comparatively, a domestic bank wire in Brazil before Pix was already easier and faster than one in the US today. I don't recall the bank fees being bad either.

The issue is that bank wires were never designed for buying lunch at the food court. They're not instant and not user friendly to set up.

Pix is alien technology next to the stuff we have in the US.

My understanding is that FedNow could become something like Pix, but implementation is voluntary. In Brazil, the central bank required all retail banks of significance to implement Pix by a certain deadline.

Visa and Mastercard are very much against FedNow becoming widely used, as it would destroy their business.

Calling the F-35 a masterpiece is rage-bait.

The primary goal of this program is not to make a plane, it's to spend $2 trillion in military contracts. As a side effect, it runs as a jobs program for engineers and its US based supply chain. Technology gets developed but with a super low ROI.

xv is my favorite image viewer of all time. I loved how it launched immediately and made it very easy to see an image or browse a folder right from the command-line. 20 years later, computers are dramatically faster and such a fundamental task has become unbearably laggy.

Tesla body work is extremely expensive. Aluminum, extensive welding instead of fasteners, substantially reduced modularity due to castings, specialized tooling just off the top of my mind.

Elon can't legally financially entangle Tesla to SpaceX due to Tesla being a public company, so his hands are tied.

Tesla is clearly benefiting from protectionism and its sales would collapse if BYD were allowed to openly sell in the US. Most people just want affordable, maintainable and reliable cars.

Because it looks like Elon recognized Tesla's inability to compete against BYD and gave up making cars. This is negative.

Since he couldn't leave it at that, he announced a pivot to a product that doesn't exist. This is also negative.

The metric for the Cybertruck is the impact to Elon's ego. Nothing about this project is rational.

No, it didn't. Order was not restored, criminals were encouraged, and here we are.

The Trump administration has gone so far down the path of fascism and crime that I'm convinced they don't simply want to be in power indefinitely -- they need it. Otherwise, the moment a law-abiding president gets elected, there will be criminal charges against all involved. And there's no statute of limitations for murder.

I believe this country will need massive investigations and criminal trials to heal. I am concerned with what happens in between, but this is reality as I see it.

I always thought it was hilarious that a company called Danger lost everybody's data. The connection to Microsoft only makes it better.

My impression is that Chinese consumer products haven't been hijacked by the "design above everything else" mindset. The priority is to make things work at scale.

American product design is obsessed with appearance and finish. Products end up costing 3 times more and functionality is degraded.

Americans are too culturally isolated from other countries and cultures to build empathy. I think Americans have main character syndrome at scale, and these comments are obvious when read through this lens.

This may surprise folks who don't live in the U.S., because Americans describe their country as a nation of immigrants and say things like "I'm Italian" and "I'm Irish" when describing their identity. Yet these same folks haven't set foot in Italy or Ireland, don't speak the language or have awareness of present-day concerns from those countries.

Thanks for the explanation -- I also had no idea what this meant.

I am disturbed that in-group code has been normalized as a way to state socially abhorrent positions out in the open, as if it offered plausible deniability.

I disagree. Reasonable would be to require a minimum salary for the job, as market evidence of scarcity.

Paying an exorbitant fee to the government will discourage the activity in the United States.

Deploying a model on an NPU requires significant profile based optimization. Picking up a model that works fine on the CPU but hasn't been optimized for an NPU usually leads to disappointing results.

Yes it's hard and also not done well. Most companies don't fund security as much as they should. At best they'll hire an occasional consultant for the purposes of compliance with a supplier agreement or industry regulation they have to meet.

No doubt, Big Tech prints money and absorbs all available talent. Boeing's troubles are surely compounded by having all the best new talent hired away from them, because Big Tech pays much better.

Elon Musk makes it abundantly clear that he has absolute control of his companies and publicly entangles them (e.g. sending Tesla engineers to audit Twitter).

X was operating in Brazil while being maliciously non-compliant, in a manner obviously directed by Elon.

Ergo, Elon is playing games to operate global companies without complying with local legislation.

I am pleased to see a government willing to put a stop to this madness, and I am comfortable with piercing the corporate veil to prosecute this bad actor who is at the root of public and consistent malfeasance. The US government seems entirely unable or unwilling to offer any enforcement.

Elon Musk is not in Brazil so action against him would be a waste of time.

Action was against X, and then Musk shut down X's Brazilian office and left outstanding debts. The supreme court evaluates that X and Starlink have same ownership and therefore Starlink (which still has local representation) is being held responsible for X's delinquent debts.

I am not a lawyer so can't comment on legality of this but it's obvious that X's stance on free speech is incompatible with Brazil's legislation on hate speech. My opinion is that X never had any intention to observe Brazilian law, and ran out of options to delay and deflect.

Alternative take: what would you expect by consistently and publicly antagonizing a supreme court justice?

Is it? Hardware engineers at Apple, Microsoft, Meta, NVIDIA, Amazon are paid much better than at Intel.

One problem with paying people significantly below market is that some of your people will be disgruntled and will optimize for working as little as possible without getting fired.

Soon enough there are entire groups whose culture aligns with cutting corners and pretending to work.