Jeffrey P. Bigham

an AI bubble can still happen even if the AI industry is doing fine -- it seems lots of investors have gone wild, over-leveraging on AI-related bets that won't come out on top, even as the industry itself races ahead

this week i learned that people have all sorts of folk theories about when they should compact the context when using AI

Helping authors improve their papers is not the purpose of peer review.

I can’t imagine anyone at all would care if comcast went out of business — too many years of terrible service, predatory schemes to extract money from customers, etc

i was not a big AI writing hater, but some of these papers i'm reviewing are incomprehensible slop.

i told y'all you should have stuck with Perl …  tech.slashdot.org/story/26/10/...

maddening, there are so many completely uncontroversial areas that we need to be investing in to maintain our leadership in science

if you were wondering what is too unsafe to ship, the proposed threshold is 50 human lives -- techcrunch.com/2026/10/20/a...

i'm starting to get a sense for how trump bankrupted a casino

imagine giving $276 million to elect republicans, and it being approximately .25% of your net worth www.nytimes.com/2026/10/12/u...

sure, but what if the slop has something to say?

I’m pretty good at writing in a way that no AI would do

<university party in 2029> a: hey, so what do you do? b: i'm a slop separator, math division www.theverge.com/ai-artificia...

i don't think this is real, quite, but it's a near future. and so, it is the case that research papers need to be bigger and bolder. claude et al do make a lot of research papers much faster, so just as we have before, we need to widen the size of questions asked news.ycombinator.com/item?id=4996...

assisted desk rejection, but for all the things.

i was resistant to the idea of the AI bubble bursting, because there is obviously huge impact in AI. but, then i heard about how every rich person in the world is scrambling to throw money into AI, and it started to seem overextended. maybe not every bubble is tulip bulbs.

do you know what I never want -- the university template page for a faculty member.

it is the age of truly personal software

let's go carnegie tech!

u.s. news now weights earnings after graduation more highly, and so MIT bests Princeton, and CMU rose 5 spots to #14 … obviously any metric is imperfect, maybe they should just publish a nice interactive visualization? more useful, would sell fewer magazines www.nytimes.com/2026/09/22/u...

Proud that our Workers' Algorithm Observatory is partnering with NYC's Dept. of Consumer & Worker Protection and the Workers Justice Project to build tech infrastructure for compliance monitoring of DoorDash, as part of a record $131.5M settlement✊🏽 Press release: www.nyc.gov/mayors-offic...

me too! www.nbcnews.com/tech/tech-ne...

with more than 50,000 abstracts submitted to ICLR last week, can’t decide if that’s evidence ML research is stronger than ever or evidence that it’s dead

The Muse app continues Meta’s trend of opting users into data collection for AI training. It also nudges you to share your bank account, email, and passport information. www.wired.com/story/metas-...

i wonder if all my carefully curated llm wiki/project folders will still be relevant in a year?

i'm still useful! or, maybe it's just being sycophantic?? :(

There are things I disagree with but there is good stuff to learn from taking the perspective of parts of the cybersecurity industry that rogue AI incidents may be best understood as security & organizational failures that let rogue behavior to turn into crisis. www.normaltech.ai/p/the-ai-as-...

EXCLUSIVE: Humans are reading ChatGPT users’ prompts to improve OpenAI’s models, and those chats can include sensitive, personal information, according to leaked internal documents and real prompts seen by 404 Media.

AI might really be out of control right now, but it's no real surprise that nobody is listening to them say that it is -- remember when GPT2 was too dangerous to release?

in designing AI futures, i think we both under and over estimate how likely folks will be to just accept the AI's judgment on higher-level strategy things. i'm already seeing people just using claude's weird dark mode purple slide style, imagine that for all the things.

the democrats should say they will give people $5,001 if people vote for them. or maybe $1 million! it's all bullshit anyway.

write your paper with AI, then use this service to make it sound like a human again, hahaha.

Imagine if you just gave mathematicians 22 million dollars

came into work, forgot my running shorts, uber eats'ing a new pair from dicks sporting goods. three computers with agents going on long-term explorations with a goal, wiggum but actually pretty confident it'll work. and so i'm still skeeting.

or am i …

all my work is done via agents now, although i am skeeting this myself, so count yourself privileged (or not).

the thing that will be wild about the anthropic and openai IPOs is that suddenly the politics of their employees will become front and center because they will have lots of money, like crypto but maybe 100x?

using AI today feels like those $3 uber rides in 2015

i'm a sloppy guy living in a sloppy world, it's sloptastic

ok, claude is a better coder than me, and it's like 1000x more persistent. lol.

as AI makes computing easier (and harder), people remain the hardest problem in computer science.

a hard part about having debates about appropriate use of AI is that people have dramatically different experiences with it … and those most against AI often have the least, so their specific arguments/examples don’t land

decided to not review for venues that i do not submit to -- baby steps toward sanity and focus. haha.

i love my chevy bolt, but getting rid of the carplay for some proprietary (worse) thing, not gonna get one of those … www.autoblog.com/features/202...

i've never been an organizer, but because of AI, i now put things into folders, because AI is most useful if you curate your content for it.

someone made a joke about how there should be a hotornot for objects, like soup, and in 10 minutes and $3 later (for the domain), the site exists … hotornot.life totally useless and not even very funny, but so wild.

tired: human language wired: tools inspired: code

smell ya' later mechanical turk www.cnbc.com/amp/2026/08/...

what kind of AI disaster would be enough to move the needle -- we have weirdly staged AIs hacking into other company's systems, AI's autonomously killing people in wars, steady escalation but nothing seems to have moved the needle. is that b/c we're confident or naive?