ericpauley

These FAA compliance cost estimates are always hilariously nonsensical. In what world are turbine AP hours being billed at $85/hr?

I was finally forced to go CLI from the VSCode extension because its treatment of sub-agents is abysmal.

The new sidebar design was unnecesary, a waste of space, and the rounded corner cutting into the website is flair with no functional purpose.

The fact that I've historically needed to monkey patch the Firefox chrome CSS (and will have to again to remove the unnecessary bar left of tree-style-tabs) would to me imply poor design.

It's actually almost entirely DNS. Also rest assured that I have definitely asked how these people get their data [1].

[1] https://news.ycombinator.com/user?id=ericpauley

Interestingly most traffic coming out of this AS was actually UDP: https://dashboard.terracenetworks.com/sources/asns/13238

Pretty benign AS as far as actual exploit or malicious crawling is concerned.

I'll know we've hit AGI when Claude finally realizes you can't just linearly interpolate pareto frontiers...

Interesting, but the article would be far more enjoyable to read if it weren’t clearly written by Claude.

This has “/dev/null as a service” vibes…

Believe it or not there are entire companies that are (to simplify, of course) just that: https://www.adjust.com/

July 11 last year I got a Crucial 128GB kit for $299. Those were the days…

Presumedly they collect that, but the vast majority of blocks are correct and not errors. This info allows them to look up the user’s request to label it as legitimate.

Guessing Law Enforcement

AI;dr

Trying to sleep straight through a long haul flight; what’s the issue with that?

AI;dr

> rocks, sediment, dirt pee, or literal poop

This is basically what dust is. It’s everywhere.

How are you potentially damaging your eyes? If it’s totality looking through them is a waste of time (you won’t see anything) and outside that range you have hours to pass them around. The sharing part has no impact.

This is incorrect. The plan doesn't apply to workspace aliases, but it does apply to workspace users using Send-As via SMTP.

The flip side of this is that GMail's Send-As feature has seen virtually no maintenance whatsoever. It's buggy-as and I expect most people (like myself) have been putting up with it out of inertia because switching email clients is a pain.

This was the last straw for me. I'm migrating to Apple Mail and cancelling workspace ASAP.

ETA: I said nothing about where I'm planning to host my SMTP server. Apple Mail is a provider-agnostic client-side application, like Thunderbird.

ETA2: Those who read TFA carefully will know that, even though this doesn't affect Workspace aliases, it does affect this feature (Send-As SMTP) for workspace accounts.

I was really hoping for this article to eventually get somewhere, but it is mostly just random anecdotes about the author’s mom using Claude.

Trademark C&D in 3, 2, 1...

Project-based learning is not a new idea. However, it is precisely the sort of assessment strategy that is being attacked by AI: students can complete many projects fully using AI without learning anything about the underlying material.

In every state I am aware of employers are required to gross up to minimum wage if tips fall short. Yours is likely the same.

This idea that the US labs are just directly committing fraud against effectively every major US organization is the most tinfoil hat thing I’ve heard in a long time. Training on excluded data would (eventually) be trivially provable. Forget loss of trust; this would make the labs a defendant to the most legally well-resourced organizations on the planet.

It’s really about incentives more than morals. I know that, being in the US, if Anthropic steals my company’s trade secrets through model inputs contrary to our policy that I will (in theory) have legal standing and enforceable contracts against them. The same is not true against DeepSeek, and especially not against “Stealth”.

"He who reads nothing will still learn the great facts, and the details are all false."

–Thomas Jefferson

I suspect that sustained reading of Opus 5's unconscionably bad prose could actually cause psychological harm. We're strongly considering moving all of our Anthropic spend to Codex/open weight models. It's a mental health decision at this point.

Two things: (1) models are far more autonomous and run for longer coherently, (2) visibility into the development process was lost in local development.

Because models run for a long time autonomously developers need to manage more agents simultaneously. At that point you basically need a tracker for tasks in flux, and compute for each agent. It also means that the interactive monitoring isn’t as helpful anymore.

Regarding visibility, developers now think in high level architecture rather than lines of code, but implementation details are still important to document. We also want to understand when agents are hitting pain points. Both of these basically require archiving agent outputs more deeply than just code changes and commits.

I'm going to chock this one up as a yet-another "X is your agentic coding hub" PR nonsense.

What are people actually using to manage semi-collaborative agentic coding workflows nowadays? We've moved from on-dev-machine coding to @claude in GitHub, but of course GitHub purposefully weakens the capability of that approach to the point that it's kind of a hack. Github/Linear have somewhat more cohesive experiences but it still feels like tooling for agent teaming is quite lacking and unergonomic.