fny

Are there any good studies or benchmarks about compressed output and performance? I see a lot of arguing in the comments but little evidence.

Everyone seems to forget Markdown was invented for humans and incidentally for machines.

Almost everyone who complains has some parser or rendering related agenda. No one cares if you can't nest asterisks inside underscores. Most of these edge cases don't read well as Markdown, which defeats the purpose.

Trump scored an own goal. Military conflict tends to hijack the front page.

I know actual Iranians--not just expats, and I know even more people who hold Iranian passports.

The NPR article also quotes people on the ground who have managed to get messages out.

The big mistake was underestimating the appetite for rebellion despite 70%-80% wholesale opposition to the regime.[0] I personally know many, many Iranians who welcome the attacks along with their families. All of the high-profile assassinations involve intelligence from Iranians.

However, no one has guns, and government-backed militias roam the streets to maintain order.[1] There is no possibility of military coup. Many officers lives and livelihoods are at stake post-revolution, and they will go to great lengths to protect it. Remember, they killed 30K of their own to quell an uprising.[2] Surveillance is everywhere online and in person.[3] One spy in ten can ruin a revolutionary group. To make things worse, there is no unification around a leader or what should come next.

If anything, this war demonstrates the tyranny and tentacles of the modern state. The well seems forever poisoned once power is lost to despots.

[0]: https://gamaan.org/2025/08/20/analytical-report-on-iranians-...

[1]: https://www.npr.org/2026/03/19/g-s1-114144/iran-voices-war

[2]: https://www.theguardian.com/global-development/2026/jan/27/i...

[3]: https://www.pbs.org/newshour/world/iran-built-a-vast-camera-...

That's not the use case. The use case is running apps from a remote Linux host as a local window. A performant VNC for specific windows if you will.

For example, you could run VS Code on that machine as a window on your Mac. A more real world example is people accessing guis (e.g. matlab) on lab clusters.

The closest set up for x11 would be to use x11 forwarding with xpra.

This is equivalent to "do people find PowerPoint to be helpful or distracting." Sometimes yes, mostly no.

In this case, I'd say helpful because I didn't have to read the article at all to understand what was being communicated.

This fine from New Mexico is about 0.6% of Meta's annual profit.

If all 50 states sue at the same rate, that'll be a 30% dent, and I'm sure states can sue for more than 0.6% too. That would be historic action against malfeasance and would send a strong FAFO single to all corporates.

Let's lobby for it.

It uses the Homebrew API and uses its own dependency resolver and linker to pull Homebrew's precompiled packages.

Claude Code was released for general use in May 2025. It's only March.

Also using PyPI as a benchmark is incredibly myopic. Github's 2025 Octoverse[0] is more informative. In that report, you can see a clear inflection point in total users[1] and total open source contributions[2].

The report also notes:

> In 2025, 81.5% of contributions happened in private repositories, while 63% of all repositories were public

[0]: https://github.blog/news-insights/octoverse/octoverse-a-new-...

[1]: https://github.blog/wp-content/uploads/2025/10/octoverse-202...

[2]: https://github.blog/wp-content/uploads/2025/10/octoverse-202...

You absolutely can see a difference. [0] The term of art is "Runway Incursions", and the stats definitely show our airports are working at the limits of safety.

[0]: https://www.buckycountry.com/2025/09/22/runway-close-calls-u...

Funnily, this started because their trying to avoid perceived insta shadow banning.

In clinical settings and situations where probabilities really matter, its a better fit.

I studied stats at Duke which is a Bayesian academy. Almost every problems come from regimes with small sample sizes. Given that Duke houses the largest academic clinical research organization globally, having a stats and biostats department with this bent is useful: samples are tiny in clinical trials compared to most big data settings.

The biggest problem with the whole Bayesian regime IMO is that as the data gets larger its selling point vanishes. If your data is big or is normal (mean-based statistics), a frequentest/bootstrapped CI approximates the Bayesian CI anyway.

Furthermore, many us work in settings where we're trying to sell toothpaste: we don't need the Bayesian guarantees that an insurer might.

Unfortunately, cops have the right to break down your door if the warrant warrants it--like having a sex dungeon.

Here's a playlist that includes them: https://www.youtube.com/playlist?list=PLcopJNj12opkoEDaj5lwP...

I highly recommend people watch video from the trial--specifically the officer testimonies. It's absurd this lawsuit was even fit for trial.

Other countries need to invest collectively in open alternatives, and AI must be considered critical infrastructure rather than a commercial venture. Building small firms to compete against behemoths will not accomplish that.

And by open I mean open weights AND open training pipelines.

The other consideration is that the kill switch is ultimately controlled by the US. The US government can easily commandeer Starlink or jail Musk, but other countries use starlink at the pleasure of both Musk and the US government.

I never understood how people use compiled languages for video games let alone simple GUIs. Even though I'm now competent in a few, and I have LLMs at my disposal, I fall back to electron or React Native just because it's such a pain in the ass to iterate with anything static.

Native devs: what are your go to quality of live improvements?

As the kids say, AI makes everyone a 10x engineer. Who would you you want to 10x?

Even though a lot of what people with agents is wreckless, they often build their own guillotine in the process too.

Problem #1: He decided to shoehorn two projects into 1 even though Claude told him not to.

Problem #2: Claude started creating a bunch of unnecessary resources because another archive was unpacked. Instead of investigating this despite his "terror" the author let Claude continue and did not investigate.

Problem #3: He approved "terraform destroy" which obviously nukes the DB! It's clear he didn't understand, and he didn't even have a backup!

> That looked logical: if Terraform created the resources, Terraform should remove them. So I didn’t stop the agent from running terraform destroy

I feel like the push by devs towards Metal compatibility has been 10x than AMD. I assume that's because the majority of us run MacBooks.

There was never a 3.0...

I hate MCP servers

That said the core argument for MCP servers is providing an LLM a guard-railed API around some enterprise service. A gmail integration is a great example. Without MCP, you need a VM as scratch space, some way to refresh OAuth, and some way to prevent your LLM from doing insane things like deleting half of your emails. An MCP server built by trusted providers solves all of these problems.

But that's not what happened.

Developers and Anthropic got coked up about the whole thing and extended the concept to nuts and bolts. I always found the example servers useless and hilarious.[0] Unbelievably, they're still maintained.

[0]: https://github.com/modelcontextprotocol/servers/tree/main/sr...

The trick I've found is to vibe libraries that do one thing well with clear interfaces. The experience becomes more like importing a package which arguable has the same cognitive debt issues described above.

Editing a one shot on the otherhand reminds me of trying to mod a Wordpress plugin.

You can have other sensors that tell you it's a screen, maybe require a Live Photo, maybe also upload to a third party service faster than generation is possible? In the end I think we'd end up somewhere like with cryptography: generating a real fake might be theoretically possible but it could be made prohibitively expensive to generate.

This is an extremely clunky and buggy prerelease, so don't try to hot fix prod from the toilet without a different mobile frontend.

Right now:

- You can't interrupt Claude (you press stop and he keeps going!)

- At best it stops but just keeps spinning

- The UI disconnects intermittently

- It disconnects if you switch to other parts of Claude

- It can get stuck in plan mode

- Introspection is poor

- You see XML in the output instead of things like buttons

- One session at a time

- Sessions at times don't load

- Everytime you navigate away from Code you need to wait for your session to reappear

I'm sure I'm missing a few things.

Isn't it a simpler solution to create some protocol for a browser or device announce an age restricted user is present and then have parents lock down devices as they see fit?

Aside from the privacy concerns, all this age verification tech seems incredibly complicated and expensive.

Don't banks/insurers/whatever have websites that are often mobile friendly?

I really hope Mario who wrote the engine that powers OpenClaw[0] gets spoils as well.

OpenClaw is mostly a shell around this (ha!), and I've always been annoyed OpenClaw never credited those repos openly.

The pi agent repos are a joy to read, are 1/100th the size of OpenClaw, and have 95% of the functionality.

[0]: https://github.com/badlogic/pi-mono