apothegm

Bulky? Costly? Software doesn’t support it? How much do they cost, anyway?

I bet they’re also absolutely awful to read/write text or code on. And the quality for 3D uses is still mediocre.

I could imagine their having an application in CAD and 3D modeling if the price were right. But the software tooling would first need to be able to support the display. And I doubt that’s a dimension (ha!) that would be worth their while to compete on (bit of a chicken/egg problem).

You’ll probably want to use an escrow service. eBay used to and probably still does integrate with a couple reputable ones.

Alternatively, sell to a reputable third party that buys and sells secondhand. For Macs, OWC comes to mind.

A robot that washes your clothes isn’t much use if you can’t afford clothes (let alone the robot) because there’s not enough work left that they can’t do to employ more than a minuscule fraction of us humans and the techno-feudal barons have no interest in distributing let alone redistributing the proceeds.

Except that obesity happens in large part because we eat too many things (highly processed junk food, sugar added to everything, and All The Carbs) that are dense in macronutrients but sparse in micronutrients.

That lack of micronutrients leaves us undernourished and with cravings (that tend to get misdirected into more junk food) despite consuming far more calories than we need.

Foods that are dense in micronutrients are precisely what we need more of.

The trouble being that the junk food is not what’s likely to be replaced by organ meats.

Surgical masks aren’t. But N95/KF94 were originally designed for exactly this sort of use case.

They’re useful if you want a job with a big company that doesn’t have tech in its DNA.

Tech companies and startups associate certs with big traditional companies that aren’t really tech savvy — they’ll either ignore certs or see them as a negative signal. The idea being that certs show you’ve memorized something but not that you know how to think.

Small non-tech companies typically don’t know enough to look for such a thing.

Because the world and the jobs we need to put food on the table and roofs over our heads work on whatever the local time is and don’t allow for waking up on standard time if your locale is running on DST?

Since climate change is making extremes worse, not just causing warming… likely more than baseline.

Doing what everyone should be doing and using the LLM for what it’s good at (natural language processing) and not for what it’s bad at (decision making). In a high stakes context like this, trusting an LLM for anything more is negligent — arguably criminally so.

All the private files on the hard drives (and maybe cloud drives depending on host) of individuals who haven’t enabled AI assistants on their devices.

Corporate data belonging to non-AI companies.

Data that exists only in filing cabinets or on microfiche.

Whatever’s on that collection of CDs and floppy disks that you still have sitting around.

The home videos nobody’s bothered to digitize.

Private art collections.

Photo prints from film cameras that haven’t been digitized.

Airplane black box data.

Personal handwritten journals.

Plenty of private dead tree correspondence.

The vast majority of grocery store receipts.

Properly stored passwords.

Proprietary code produced by companies that don’t use AI or use only self-hosted models.

Scientific data that hasn’t been published or analyzed with AI, including plenty of older data.

Annual reviews predating LLMs.

Information exchanged during phone and video calls that aren’t recorded or haven’t been submitted for transcript generation / summaries.

Most board meeting minutes.

Anything handwritten that hasn’t been scanned.

The contents of many private online communities.

A lot of classified information.

I think that’s enough for now.

Starts talking about LEDs and totally misses LCDs, which made heavy use of 7-segment displays and were ubiquitous for a few decades in the early years of consumer digital electronics.

I might be off by a factor of 2 in that estimate, but not much more.

Most of our communication is via chat. Then there are ticket specs and comments. Project overviews and status updates. PR descriptions and comments. Documentation. ADRs. Retros and incident reports. Meeting agendas, notes, and writeups. The occasional email. LLM prompts. Last week I spent 5k words in a few hours just breaking down a large planned project to ticket-level bullet points so we could give an appropriate estimate for roadmapping. (Which turned out to be crucial, because it revealed that the requested timeline for a pilot was too short by at least a factor of 2 even for the most descoped version the stakeholders would accept, and we had the receipts to convince the CTO.)

I’m definitely not spending most of my days writing novels, but all the sentences add up.

Okay? This article is about the campaigning for change part. It doesn’t say you shouldn’t make those tweaks. Just that you shouldn’t have to and we should change that.

LLMs? No. 85% correctness and avcuracy — even 95% — is just not going to cut it in that context.

Easily. I probably generate 180k words per quarter at work and another 180k for personal and social media purposes. Maybe 180k/month, even.

Best performing headphones I ever had were the ones with the foam covers that came bundled with a particular model of Walkman. Kept those together with duct tape until the electronics finally gave out.

I’d be more concerned about the refusals than the refills granted.

It’s a US holiday.

Which would in turn imply that markets cannot be simultaneously efficient and competitive.

Wouldn’t (1) be the point?

Composer writes the worst, stupidest, most naive and straight up brains-dead code you could imagine. Fast and cheap is about all it’s got going for it. I mostly use it for “sort these lines alphabetically” and stuff that’s a smidge too complex for regex find/replace.

That last one in particular is turning AI into an enemy for a LOT of people, and it’s hard to blame them.

Why is Grok just so sexualized?

I would think the answer to that one would be obvious, tbh.

For those of us without a WSJ subscription… what is it?

I think you mean (1-(1-prob[success])^numAttempts).

Shocking. Totally shocking.

Water usage concerns in a desert, perhaps?

The learning curve is arguable, but the discoverability is essentially nonexistent — and that massively affects the learning curve.

Yes. The trouble is that to be able to not starve to death while you’re looking for a new job, you have to sign away a ton of rights, basically including the ability to sue for wrongful termination or expose wrongdoing. It’s a situation with an extremely unequal balance of power/leverage.

For Gen Z the new reality is for other people to make money off your every action, squeeze you for every cent, and not be willing to pay you enough to live on no matter how hard you’re willing to work — or even to employ you at all, in favor of robots and agents. Getting worse every month.

The hope is that the government will tax the absurdly wealthy extractors and offer a UBI so that everyone else won’t have to starve in the streets or bring out the guillotines.