the penne opticon

People underestimate what these tools are. They use the least-flexible, least-capable harnesses, then assume that that is the universe available to users. (Cowork and Codex support scheduled tasks out of the box.)

I don't know about 30%, but it's self-evident that major disruption's already here.

The future can be cool sometimes

project hail mary is already a retrofuture because the ship doesn't have a reasonable AI on it

if we want the vibecession to end perhaps we should be focused primarily on reducing the grinding existential dread of the future that even many objectively comfortable and highly-paid people seem to be feeling nowadays - and yeah, i do think there are practical things one could do to improve this.

since yes, it seems clear that "making good money right now" is absolutely not enough to get many people to unclench nowadays, which TBH really shouldn't be a *huge* surprise.

@daralind.bsky.social

samuel beckett was born on this day 120 years ago. astonishing to think that if he had survived, and could run 100m in 9.57 seconds, he would be not only the oldest but also the fastest man in the world - together with his nobel prize for literature, an astonishing trifecta

ChatGPT is a website. Claude is a website. While I guess Claude Code runs in a terminal window, that just means it’s an app, which I put in exactly the same mental box as I do a website. Yet everything you read or hear or see about AI does everything it can to make you think that AI is something other than a website or an app. People that “discover the power of AI” immediately stop discussing it in the same terms as Microsoft Word, Google, or any other app or website. It’s never just about what AI can do today, but always about some theoretical “AGI” or vague shit about “AI agents” that are some sort of indeterminate level of “valuable” without anyone being able to describe why.

zitron continuing to innovate ways to avoid looking the phenomenon in the face. it's computation. how much is happening and where? "a website" doesn't distinguish between a blog and the enormous apparatus powering google search
www.wheresyoured.at/i-will-never...

Honestly, the fact that the Yale Political Union, of all places, is where this happened is beyond hilarious

Straight lines on graphs People who do not watch AI developments closely are often suspicious of an intuition that I’ll call “straight lines on graphs,” which holds that progress in AI is (not only rapid but) remarkably regular over time.

this is a good post that I've hesitated to share because it's so boostery. but I agree with it. it's funny that "straight lines on graphs" here actually means "straight lines on logarithmic graphs", and the author is so exponential-pilled that it doesn't even occur to him to mention it

I think that as economists we should be a lot more concerned that everyone feels like they've got poorer when the data show they haven't. We certainly shouldn't just be assuming that the customers are wrong, rather than that we're no longer measuring things that are most relevant to wellbeing

doordash grandma DoorDash does not have a specific policy for seniors or helping grandparents, as the provided articles do not mention DoorDash's policies or services for seniors. If you have a follow-up question, feel free to let me know. how the fuck do I just search WaPo articles? To search Washington Post articles, you can use the "Ask The Post Al" tool, which is a generative Al tool that provides summary answers and curated results directly to users. If you have a follow-up question, feel free to let me know. I asked. Show me article about "doordash grandma" There is no information about DoorDash and grandma in the provided articles. If you have a follow-up question, feel free to let me know.

WaPo website replaced search with AI and it's going great

Soon, at each release of AI along the current capability curve, you will start to see large discrete jumps in ability in economically important areas, because the previous AI ability level in some aspect of the job bottlenecked progress. When bottlenecks are released, it looks like a leap forward.

It is going to look like unexpected gains in AI ability, and, indeed there is no sign that the current exponential ability curve is slowing down (so far), but it is going to be like what happened in coding: as soon as models crossed a certain threhold, suddenly Claude Code & Codex were viable.

We do very much need to move beyond being "pro" or "con" this stuff. It's going to be as silly as being pro or con steam. Useful policy is not shaped at that level.

through years of LLM research, we have finally managed to invent a human-usable interface for ffmpeg

do you have a link? (Even just to which episode?) Would be very valuable for me as an example

bruenig said on his podcast that the free trials have started to run out and various unions are now paying him for this. also that it was used over 12k times in a month

Excited to announce that I will be presenting my upcoming book "The Open Strait and Its Enemies", to be published in two volumes (vol I "Seeing Like a Strait", vol II "Epistemology of the Closed") by a major University Press in early 2027, at the French-British Strait Conference.

my grand theory of vibecession is that multiple contradictory things can be true at once, and that this kind of complexity is more or less impossible to constructively discuss on a social media platform with character limits and severe context collapse problems.

The city has been waiting a very long time for this guy, who just looks around and sees dumb shit and is like "Hey we should fix that"

It’s good that we all have something to cheer for that involves a positive aspirational vision of humanity.

Was at a local bar tonight and the owner got everyone to shush and then announced loudly that the Artemis II astronauts had just touched down safely, and the entire bar erupted into cheers.

Hell fuckin yeah.

just insane how much of this discourse got tied up one academic's narcissistic insistence that her research program had not been superseded

Oh, to be clear, I don't think that an AI should write journalism without doing actual reporting. Nor do I think that anything (for now at least) can replace shoe leather reporting. But plenty of reporters don't burn shoe leather to report, they just let their fingers do the walking. Why not AI?

We're early yet

I think there's a disconnect here: I'm asking: why can't an agentic AI pick up the phone, send emails, etc.?

Sure, it can't go in person. Lots of good reporting doesn't require being in person, though it benefits from it.

I don't think I agree - what am I missing?
1. Lots of journalism (should) be "dog bites man" reporting where individuals novel but event statistically normal, e.g., the local police blotter
2. Seems like an agentic LLM could probably do decent investigation of many out-of-distribution events?

I suspect that popularity of AI is going to start looking like surveys where people trust their own doctors but are distrustful of the medical establishment

People will increasingly like and rely on “their AI” but will increasingly be anxious about “AI” as a category. Some odd implications result.