Info-dense, completely unreadable README.md file. Only the line "No Puppeteer, no Chromium IPC, no `page.screenshot()`" actually allowed me to figure out what this is about: creating video demos of your website (and probably other software as well) with an AI agent, rather than going through the conventional video-editing process.

It's cool that so many people are building useful tools with LLMs, but for the love of ones and zeroes, people! PLEASE write your README by hand, instead of handing it off to the LLM with no human involvement! You know how unreadable LLM prose is! Why do you make that the first thing people see when they see your repo?

I will accept the "bad things" if and only when Sam Altman exposes himself to those bad things first e.g., living next door to an AI data center, being subjected to 24/7 surveillance, and having everything he says and does scrutinized for illegal activity, ticketing and jailing him appropriately.

Until that happens, no. Sam Altman is here telling us to accept the cruelty of his actions and his company so that he can be more wealthy.

> I don't know what you mean by "trusted",

It means "hi spending approver, I'm going to add $100 to our CF account" instead of "hi accounting+management+security, please initiate the process of evaluating new third party vendor Foo for use in my project, I hope we can get it approved and integrated into SSO sometime next month".

Those “small” models take up 25 gb. On an entry level mac that’s a double digit percentage of available space. It is very rude of apple to claim that much space in an OS upgrade and offer no easy way to get it back.

Some great soul said something like "God has given everything to meet your need, not your greed". As long as the greed continues to increase exponentially, no amount of innovation is going to support it.

The amount of energy and material wastage in developed countries is enormous and abominable. Large screens being on 24x7 without anyone watching, cities explicitly designed for forcing the use of vehicle, consumer life-style requiring huge supply chains, powerful cars carrying just one person, ... and so on.

Business drivers will continue to drive the greed of people and nations and act against any measures to fix the problem. A business knows nothing about climate or human good.

This is not an accurate description. The first identified patients were people working in or living near a "wet market" where (among other things) various wild animals were sold. It's plausible that someone was infected somewhere else and the market was only the site of a "superspreader event" (rather than the earliest jump from some other animal to humans), but further details are speculative, informed by indirect phylogenetic evidence rather than direct forensic evidence.

It's not unusual/unexpected that there were labs working on coronavirus research in several major Chinese cities: the 2002 SARS outbreak, which began in southern China and caught researchers and public health officials off guard, led to significant investment in the study of coronaviruses.

> The novice became suspicious and said: "Master, is all this 'Tao of Backup' stuff just so you can sell more copies of Veracity?"

> The master said: "Now you are truly enlightened."

The firmware situation is mixed. The ECU hardware is a standard TAG700 from McLaren. Basically these are software-defined ECUs where you can do whatever. However, FIA has some rules on what you're allowed to modify and has some "compulsory" software in there for monitoring and competition regulatory reasons. It appears there was very limited testing of this module in wet conditions which seems to have affected a lot of vehicles. (I work in Motorsports)

Bright lighting is typically an artifact of the commercial property insurance policy in effect (namely, "Protective Safeguards"). It's asking you to demonstrate that you've done enough to create a barrier towards theft/loss.

It doesn't have to be really bright lighting but that's the easiest thing to satisfy the underwriter's request, and as a result, the cheapest to implement.

Broad over-generalization is code for "no article"

Why would a movement which started as nerds buying mosquito nets to stop malaria be "a bunch [of] post hoc justifications for bad behaviour"? You say this like it's an established fact or at least a common argument, yet I've never heard it before.

With the caveat that I have not watched the video, the title itself rubs me the wrong way. Intent, quality, and artistry as I understand them are not things intended to "scale." You would not, for instance, as a musician, think "I made an album, yes. But how do I scale up to ten or a hundred albums?" You may well think "How can I scale my audience," but that is a different question entirely and not a matter of intent, quality, or artistry. Scaling art is, to me, a wrongheaded notion to begin with.

Yes but draw an arbitrary line around any group. Call them HNdrathals if you want. Define them at one instaneous snapshot in time as 100% conserved gene corpus. Then watch generation after generation, as some of the HNers (against all odds) interbreed with the surrounding populace. After one hundred generations, should we expect many people to be more than 2% HNdrathal? Yet did the HNdrathals ‘die out’?

Aside from the incredibly low quality of the data, this jumps out as being vibe coded due to the unbeleviably stupid text.

"Rain as a surface; Height is millimetres of rain per day. Nothing else is encoded."

Why does AI generated ridiculous sentences like these?

> That's not going to reach AGI,

It has not been even 4 years since ChatGPT hit and LLMs + Transformers + Whatever they do has gotten us to solving millennium problems.

4 years ago, a program that could create photorealistic pictures, talk to you in any language of the world and solve the hardest math problems that we know, we would have called it AGI.

Now I don't know if what we have is AGI or not but I do not understand how you can see what has happened in the last 3 years and say "it will not get us there" no matter what "there" is.

Actually you are wrong. They are only waiting for them to get to the point where they can't push LLMs any further via training and then lobby Governments to introduce safety regulations that they'll adhere to, thereby using safety itself as a moat to block competition (including open weights from other countries, notably China). That's exactly why they are using China as the bogeyman and you keep hearing hysteria about AI labs not pursuing "safety". Safety is going to be the moat that will lock AI into these Big 4 AI labs.

Man that was as a different ethos back then…”Draper realized that his ‘learning curve was flattened out’ in venture capital and detoured into public service”. Subsequently he funded 280 non profits.

The A16z team of today is probably deeply confused reading this article.

"I have nothing to hide"

Pro-surveillance people were not born one fine day. They were conditioned for a decade making them believe sharing everything about their life is normal.

"What's on your mind?"

All of these projects targeting low spec systems and "100 tok/s" are the same 2 bit quant without much else. Conveniently none of them include any mention of accuracy in their published numbers. 4 bit is the floor.

One should be careful with vibe-written / vibe-aided articles here. Who knows what is actually supported vs what sounded okay and the author (who is not a clinician) then signed off on.

Even the core idea, 'rejection sensitivity' is acknowledged not to be a diagnostic entity: so then how can we make definitive comparisons? Maybe clinical gestalt... but the author is not trained to see and evalaute patients.

Everyone should read the above. Cringely had so many stories where he seemed to be part of, or present to witness, significant events in the computer industry - after a while you become suspicious that one person could really do/see all these things. Like he was a 'Forrest Gump' of technology.

Every bubble goes through this phase just before it pops. Pre-implosion Twitter had tons of folks doing useless “research” projects, WeWork had folks studying furniture design’s impact on wellbeing, and so on.

When you’re more worried about this sort of stuff vs generating profit then the history books strongly indicate that’s usually not a good omen of what’s coming next.

As far as I know there are still some things "the platform" actually won't deliver; for instance I'm fairly certain there's no searchable combobox that's fully accessible. One ~must leave the platform to give users the experience the expect.

That means "the platform" itself is actually teaching people to work "off platform" and arguably will always have gaps as functionality and expectations evolve.

Admittedly, the habit really ought to be 'I need to make a combobox -> does the platform have what I need? -> Research, evaluate, test -> Otherwise, make it' but I don't begrudge people for skipping the middle steps.

Yeah, you can't just implement it as a pure "stop all services immediately once I hit a set amount"

It needs to be more like "don't allow spinning up additional services after you hit this amount", although that still allows you to go over the limit by a lot, since most services are billed hourly.

It really is difficult to implement a spending cap that doesn't risk shutting down important things.

Here's a little cheat sheet for discrediting anyone who warns about AI:

* If they worked at an AI firm, say "they're a hypocrite"

* If they didn't work at an AI firm, say "they have no idea what they're talking about"

There is a command in oh-my-pi called "/omfg <problem>". You explain what is wrong with agent's response, and it writes a hook to make sure that the problem doesn't happen again. It then re-runs your previous prompt to make sure that hook is triggered, and if not, it rewrites the hook to make your previous prompt trigger the hook. Then each next agent's response is checked by the hook, and if it is triggered, the agent receives feedback on what's wrong and what must be done differently.

That is a technical solution to a problem that has a far cheaper, and less easily abused solution: don't put up cameras at all.

Building a system with "safeguards" in place that require constant oversight is destined for either regulatory capture of the oversight, or slow dismantling of the oversight.

E.g. warrants -> fisa court warrants -> Xkeyscore

And not even cold hard cash but “Cloudflare credits.”

What an offensive slap in the face.

The detail that got me: both on gurneys holding hands, and the mom crying tells him "don't worry about a thing, I've got a Genesis concert and I'm not missing it". They both made it to the concert. He still has the ticket stub

>We trained Kolibri with abstention data and with our Merlin-Arthur protocol. As a result, it is trained to say "I don't know" when the answer isn't in the context.

https://aleph-alpha.com/en/blog/bounding-hallucinations-merl...