sneak

Still an open question: How does S3 Glacier Deep Archive do it? Tape has been theorized but I’ve also heard arguments for very slow flash…

The fact that Linux is a huge inconvenient waste of time to use for normal tasks has nothing to do with brainrot. You’re moving the goalposts.

You must realize there are things like 3blue1brown, Veritasium, and Numberphile on YouTube, right?

I have a Tesla 3. There’s no good way to get a lot of power out of the car as an end user except via the wheels. You can get like ~1kW of 12VDC via the dc-to-dc computer/accessory system but even that is pushing it.

I am saddened by this. I’d like to be able to access the large amount of power in it in a pinch.

There is also no good way to charge it as an end user from a DC source like my large 48VDC solar battery system. I end up having to invert the DC to 220VAC which then charges it normally (involving rectifying it back to DC for the onboard batteries) - an unnecessary and inefficient double conversion.

You don’t have HDR and can’t pair and then send audio to bluetooth speakers via only the remote in 100 seconds. These basic things are table stakes. It does not “just work”.

Just get any old TV and use the HDMI input and don’t connect it to the internet. The fears of it phoning home anyway even when you don’t configure wifi are overblown (almost no TVs can do this), as are the privacy implications. Nobody cares that you watch Severance and MCU movies. It’s sort of like people who tape over their laptop camera but leave the mic alone, as if 86400 photos of your face are somehow private and sensitive but all of the conversations in the room are not.

You know why. Nobody wants to sysadmin their “friendly Linux distro” just to watch shows. Kodi is amateur hour compared to tvOS. Nowhere do people care more that it “just works” than the thing their TV remote talks to.

I write a spec by speaking, not much thought given to grammar and sentence structure. I over-describe concepts, over-describe functionality. A lot of repetition of details and concepts. The audience is not a human. Whisper converts it to text.

Lightly manually edit for speech rec failures.

LLM takes that and converts it into a much much denser, much more structured technical spec in markdown. This is issue 1 in the repo.

I edit this directly extensively, adding detail, removing misconceptions, adding sections with more dictated text.

LLM takes this and further densifies it, closer to an actual implementor’s spec. I carefully note what got added, removed, edited in this step.

I edit this carefully and extensively again. Some removals get added back, some additions get removed. This is the first commit and closes issue 1.

I then have an LLM write an implementation plan and basic project scaffolding. Makefiles, smoke tests, logging, test framework, DI. I edit the plan, and the plan is the second commit. An LLM uses both plus scaffolding plus my repo policy documents (global) to write a README. We’re still single threaded.

An agent files the first n top level implementation tickets. These refer to and sometimes quote any interface contracts in the spec or implementation plan. These usually map 1-to-1 to a module/directory. These sometimes have stub commits to create the module directory and commit interface definitions. Some larger modules have a README in their module directory if their purpose or structure or inputs or outputs are complicated or non-obvious.

Parallel agents write commit level issues that are the task items for each top level implementation ticket. One agent per module.

Then, spawn parallel agents with detailed global instructions, the full spec+plan, and the full readme, and the full code from the module it is working in, plus the issue text. Its output is a single commit on the feature branch and a PR that closes out one feature ticket.

This spawns an adversarial reviewer which has detailed agent instructions as well as our code style guides and repo policies. The scaffolding includes tests and linting and checks. Any failures spawn rework agents until the feature branch is up to snuff.

I review the feature branch diff and merge it, having read the review and rework comments on the PR. I frequently add comments about the implementation (“don’t do x, do y”) and add WIP: back to loop it back for more rework. I don’t see anything that isn’t green, formatted, passes linting, passes tests, and a (dumb) bot thinks is ready to land. Most of the time, it is. Sometimes, it did it the wrong way and I correct it.

The more modular the code, the easier it is to parallelize (obviously). It is ridiculously easy to build large software quickly now, provided it is sufficiently modular so the implementation doesn’t have to cross modules very much. The detailed spec and implementation plan up front helps a lot with this. It’s a lot like writing multithreaded programs really.

Any bugs discovered are tackled with strict TDD, one commit with a failing test, one bugfix commit that must be green and not touch the test.

Individual modules must conform to good contracts, meet best OOP practices, and fit into context. Too big and they get broken down further. Context size is actually a useful constraint for code structure here.

This workflow, the code standards, the styleguide, the linter settings, the fmt settings (obviously only applicable to non-Go as the Go fmt is not configurable), everything is global and stored in another repo and is vendored into each AI managed project repo. It has a fair bit of “I shouldn’t have to say this, but…” stuff (in the style of “fire codes are written in blood” - each warning sign came from experience).

My results are quite good. It’s as the saying goes, much like managing a team of junior devs. It is easy to end up with a LOT of parallel implementors, reviewers, reworkers, and tons of CI, but it’s still so much vastly cheaper than humans that it doesn’t matter.

Someone will make a LOT of money wringing out the inefficiencies from this workflow, because I think the “agent as replacement for third-world contracted junior dev” model is the one almost everyone is using right now.

> one guy won’t have the personal bandwidth to spend so much that the “unlimited” part actually matters,

Tell me you don’t use agents effectively without telling me you don’t use agents effectively.

What toxic waste do you think a datacenter produces? Where do these toxins come from, in your mental model?

Apple makes security updates available for the last few major OS versions, and even backports really bad ones further back on a case by case basis.

The M series MBAirs, and ofc the M1 originally, are hands down THE BEST laptops ever made. I have a maxed out MBP ofc but the Airs are unbeatable quality and price/performance.

I think the M1+ MacBook Airs will be regarded eventually like the original Macintosh originally the iPod, or the introduction of the original Air and the creation of the product class that came to be known as “ultrabooks”. It really changed everything about laptops to have a fantastic keyboard, speakers, screen, trackpad, hinge, and 12+ hour battery life… for $1k.

That’s what even low spec computers used to cost, not even adjusted for inflation. I don’t think spending $200-400/month on a hardware budget that enables me to earn 50-100x that is that onerous, we just think “six grand for a laptop???!?” But it’s ridiculously cheap for what it is. We wouldn’t be similarly shocked at a $6k car, for example.

“nation state actor” makes it sound unlikely. They sell boxes to many different governments, including the FBI. If you ever cross an international border you should have a modern device running the latest patches.

Cellebrite.

Most people don’t know how much RAM is in their laptop, so no.

People care immensely about size and weight, especially perceived size.

When it clicked that you meant MRNA the ticker symbol, it also just clicked for me that the name Moderna is a play on mRNA. Not sure why I never noticed that before.

Like the FedEx logotype arrow, now I will never unsee it.

“simply go to another platform” stopped working years ago when Apple took it upon themselves to decide which platforms you are allowed to install.

RIP Tumblr

The internet is not what’s being censored. The three platforms that everyone on the internet uses exclusively are being censored.

People then say, well, they can still access other websites! Everyone has a web browser!

Unfortunately the way people disseminate those URLs is via the main censorship platforms, and there are dozens or hundreds of domain names that form URLs which you can’t DM on Discord or Instagram, for example.

The internet as we know it is controlled centrally by Meta and Google and Apple and Microsoft and Amazon - what is done at the network layer is largely irrelevant.

The ownership class, and thus the courts and lawmakers actively protect the police, which function as an armed gang to enforce the status quo for the ownership class.

This idea that this is an outrage or will be remedied in courts is a silly one. This is the system working as intended. The people in a city who decide which laws get made want at least one guy who will come and punch you in the face for protesting or making noise or whatever.

First off, that's not true. Secondly, even if it were, it's not bribery - the only way to get the type of access the SRD offers is via this route, so it's an offer of specific terms for specific privileged access that is not otherwise available.

When I quit, I was taking 140-210mg/day (ie 2-3 70mg).

The major physiological consequences didn’t rear their heads until years later, after I had not touched it for 5+ years. I was lucky.

How much of watching fox news, eating big macs, golfing, and cavorting with natalie harp is counted in those “many hours he must be working”?

I can’t imagine he actually works that much. His output isn’t really there. We already know he doesn’t read the daily briefing (reports suggest he never reads anything at all) and it’s pretty clear 99% of what his admin does is the result of legwork from people who are not him.

CSS was done a long time ago. Stop changing it.

That’s for maxxed out knowledge work. Most of running a business is simply doing mid grade knowledge/admin/management work in a responsible, cares-about-the-outcome aligned fashion, which you don’t get from some $50k/y admin person or middle manager.

I would say that less than 25% of a startup CEO’s workload requires the “well-rested, highest quality thinking” that you describe. A lot of it is simply quantity and determination. The output of “having grit” is not that you are suddenly smarter - it’s that you are less willing to stop.

This is neither of those things. This is a bugfix not a feature and saying “this was fixed in build x, can you please retest and confirm” is not a timeline.

> I can't comment on known issues or on whether or when a fix might ship. Your report is the right channel for the issue.

> As new builds become available, please retest and add the results to FB25133383. If you have access to the macOS 26.7.2 RC or the 27.2 beta, results from those builds belong there too.

Waste your donated time and effort testing against each new build, even though we won’t tell you if we even touched that subsystem between them.

If I were this bug reporter and got a response like that (after making a perfect report including a minimal reproducible test case) I would never donate another free bug report again.

From when the statement in question was made, obviously.

I’d take the other side of that bet. $10k USD says that normal consumers are owning and driving Tesla Roadsters within 10 years.

If you believe what you write, it’s easy money for you. Deal?

Who recruited, hired, and directed the point people at his companies?

I never claimed he’s out there wrenching on starship. The claim is that he builds companies that do spectacular things, which is why his spectacular claims for the future achievements of those same companies are not taken immediately as fantastic as they would be if spoken by any other person alive.