There is a financial incentive to make the search results worse. (More searches, more ads, more money.)

There is no incentive for adding false positives to lists of malicious websites.

Sure, until their "smart filters" start considering GCP-hosted websites as pre-verified and small self-hosted websites as malicious. You know, like they have been doing with email?

Chrome is big enough that a website owner can't afford a false positive on their malware list, just like they can't afford to have all their email end up in spam for all Gmail users.

Due to their near-monopoly Google also has no incentive to avoid adding false positives to their blocklist - provided they don't accidentally block high-profile targets. And if a CxO is screaming over your shoulder that your website has been blocked, arguments about "false positives" aren't very compelling: they'll just demand you move off the "shitty basement provider" and switch to "proper hosting, like the Google Cloud"...

Permissions scoping

Then they attempt to download the missing tool or write a substitute from scratch. Am I the only one who runs into this??

I've used stuff like this for a hobby project where "effort to write it" vs "times I'm going to use it" is heavily skewed [0]. For production use cases, I can only see it being worth it for things that require using an ML model anyway, like "summarize this document".

[0] e.g. something like the below which I expect to use maybe a dozen times total.

Main routine: In folder X are a bunch of ROM files (iso, bin, etc) and a JSON file with game metadata for each. Look for missing entries, and call [subroutine] once per file (can be called in parallel). When done, summarise the results (successes/failures) based on the now updated metadata.

Subroutine: (...) update XYZ, use metacritic to find metadata, fall back to Google.

And that's probably the opposite of being school-shaped. School is hellishly boring if you're ahead of the curve.

That's a specific school problem. I think being school-shaped is not about being bored, but more about being willing to do tasks on a schedule and can learn a lot of material through a lecture style.

The pdf linked seems to disagree with your assessment, at least for Tao.

Agreed. But generally it very much depends on the school and the effort of those in and around it. Terrence was very fortunate to have parents who supported him and likely lobbied for his unconventional high school/primary school split education, and equally fortunate that his schools were able and willing to accommodate him.

It doesn't say 1k per day. Not saying I agree with the statement per se, but it's a much weaker statement than that.

"If you haven't spent at least $1,000 on tokens today per human engineer, your software factory has room for improvement" - how exactly is that a weaker statement?

My read of it was "by today", aka cumulative. But you're right that it can also be read as "just today". The latter is an absurdly strong statement, I agree.

I would love to see setups where $1000/day is productive right now.

I am one of the most pro vibe-coding^H^H^H^H engineering people I know, and i am like "one claude code max $200/mo and one codex $200/mo will keep you super stressed out to keep them busy" (at least before the new generation of models I would hit limits on one but never both - my human inefficiency in tech-leading these AIs was the limit)

Tokens evaporate when you have Agent Swarms. Have you tried Claude Code Teammates, for example?

https://code.claude.com/docs/en/agent-teams

Only for a few days - but going from $200-400/mo to $1000/day productively seems like a huge stretch.

Also the eat tokens may be compared to single-tasking - when agent swarms move faster, I need to come back to that task sooner, slowing down the multi-tasking that allowed me to use a full 20x max subscription... so the overall usage once that is taken into account is smaller.

It's probably just prompt based. Actual fine-tuning for these kind of use cases is getting less common than it used to be.

Likely. You can go into Nano Banana or ChatGPT right now, upload a pretty architectural rendering, and tell it to make it look old, weathered, winter, etc and it will come out looking very similar. Give it an example to really dial it in.

Probably, but with a small sample size like that, they should probably be taking the uncertainty into account, because I wouldn't be surprised if a lot of this variation falls within expected noise.

E.g. some binomial interval proportions (aka confidence intervals).

This was me, and now I have horrific back pain almost every week. Fix what's broken before it breaks you.

Depends if it's sufficiently transformative or not.

That just means it’s then subject to its own copyright. It doesn’t mean that the derivative work is also exempt from the terms of the original copyright.

For example, you can use a sample in a new song. And that new song can by copyrighted. But you still have to seek permission from the copyright holders of the sample to use it.

Fair use is the only time it’s legal to use another copyrighted piece without consent. And the rules for fair use vary from country to county.

Seek? In the grand scheme of things asking forgiveness only applies if you're going to not be that transformative and something like YouTube's automated copyright strikes might affect you. "Ask Forgiveness" is often a better option.

Fair use is a defense, not a requirement - You don't need permission to claim fair use; it's a legal defense if you're sued Seeking permission can backfire - Copyright holders may deny permission even when fair use would apply, creating unnecessary barriers.

This is especially true for parody and commentary.

> Seek? In the grand scheme of things asking forgiveness only applies if you're going to not be that transformative

That’s not why I said.

> something like YouTube's automated copyright strikes might affect you. "Ask Forgiveness" is often a better option.

Now you’re talking about platform quirks rather than copyright law.

> Fair use is a defense, not a requirement

Actually “fair use” is defined by law in a lot of counties. It’s not a defence, it’s a legal right.

> You don't need permission to claim fair use;

That’s not what I said.

> it's a legal defense if you're sued Seeking permission can backfire

It’s spelt “defence” (with a C not an S) and the point is you seek permission before distribution, not after

Not everything in life follows the “ask for forgiveness not permission” rule ;)

> Copyright holders may deny permission even when fair use would apply, creating unnecessary barriers.

Copyright holders cannot deny fair use in jurisdictions where fair use laws apply.

The difference between Seek and Ask Forgiveness in the situation outlined is that Seek lays out costs before hand and generally they are minimal, and Ask Forgiveness can determine costs at the will of the person sampled or remove the work completely from circulation.