- 7 Posts
- 515 Comments
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
7Ā·22 hours agoAnthropic pricing today, IPO tomorrow. Whoās excited everybody? (Itās gonna be a shitshow no matter what happens)
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
3Ā·23 hours agoand I did read the whole article. Itās mostly just forecasting and modelling
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
4Ā·1 day agoAnd why donāt you want your moral credibility questioned, Coxon? Because everyone would realise youāre full of it if they did?
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
6Ā·1 day agoJust saw OPās username. āNotkilleveryoneistā aka the same guy behind the āAI Black Death in 2 Years posted in 2024ā guy. Why am I not surprised
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
7Ā·1 day agoStandard fare Anthropic business
say, the day AI gets big and scary is always happening in the next six months, isnāt it?
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
5Ā·2 days agoAnother tech outage-this time at Apple
So trying to make a list here: FAA, United Airlines, Outlook, NSAnet, Chase Bank, Apple and WorkDay
all experiencing outages and technical issues. The birds of āshove AI into everything conceivable while firing the knowledge human workersā have come home to roost
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
6Ā·2 days agoThatās my thoughts precisely. Neglecting food regulations gets tons of food recalls, neglecting health gets measles outbreaks and other outbreaks, thus neglecting tech gives us system failures. Plus all the funding for critical infrastructure has been slashed and people fired. Itās really not that surprising
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
5Ā·2 days agoA shitton more outages have happened today
It appears that firing most of your human staff, replacing them with AI and then slashing funding to critical infrastructure has consequences. Like the insane amount of food recalls that happened this year, and all the measles outbreaks
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
5Ā·3 days agoAnother Jacob Coxon interview has dropped and it is noticeably worse than the previous
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
3Ā·3 days agoPulling some stuff out of the comments:
āWell, if the Anthropic and OpenAI pull in investors the way SpaceX did, this could be a path to reducing wealth inequality that is faster and less controversial than taxing the rich. SpaceX peaked at 226 in June. Now it is 136.ā
I have no fucking clue what this is supposed to mean. Is it more āAGI will give us UBIā bs or are they suggesting that the reduction of wealth inequality will be the rich loosing their cash from a bubble burst?
āLess than four years ago Nvidia was $15 a share. After all of the crashing so far it is $150 a share. People today are paying $150 for it. The bubble will burst, but a lot of folks still seem enamored with tulips. Trump, who still has almost two months of relatively free reign, called for reinstating Huang as Nvidia CEO. Whether he will get behind Anthropic is harder to say, unless Amodei proves to be having an affair.ā
I mean Trump also wanted Tim Cook to be the Nvidia CEO so I think heās just saying whatever. A lot of folks are enamoured with the tulips, but more are also becoming skeptical of this business model. And those tech stocks wonāt stay up forever if Nvidia keeps falling.
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 22nd November 2026English
5Ā·3 days agoI mean we still have Anthropicās IPO
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 15th November 2026English
2Ā·4 days agoJensen Huang discussing a comeback to Nvidia with his supporters
Mate, YOU RESIGNED. None of this āuhh actually I wanna go backā bullshit. Iām sure Nvidiaās board would love to take back the CEO whoās getting investigated for violating export controls
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 15th November 2026English
4Ā·4 days agoif we word salad hard enough weāll manifest the Robot God
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 15th November 2026English
2Ā·4 days agofull article about it here apparently over 6800 Nvidia employees have signed, but then again no user validation so we have no idea what the real number is
lurker@awful.systemsto
TechTakes@awful.systemsā¢Stubsack: weekly thread for sneers not worth an entire post, week ending 15th November 2026English
6Ā·4 days agoPeeked back at their blog to see if I could find anything about it. Text dump incoming:
The model with Danielās latest median parameters predicted that when an Automated Coder (AC) was developed, AI software R&D uplift would be ~2.7x and the pace of progress would be accelerated from the 2024.0-2025.5 level by ~1.25x.
So naively, if weāre at a 1.7x AI-attributed pace of progress multiplier, we might think that AC has been achieved (as of Model 4).
Based on the METR report and Anthropicās Fable 5.2 system card, it seems unlikely but not totally implausible.
In the Fable 5.2 system card: Anthropic says: āwe have been using Claude Mythos 5.2 extensively in our own day-to-day research and engineering during the pre-release period, and it does not seem close to being able to substitute for our Research Scientists and Research Engineers.ā This is the strongest evidence against AC already. On the other hand, METR surprisingly doesnāt rule out full AI R&D automation: āWe are uncertain, but think itās unlikely that [Claude Mythos 5.2] can fully automate the AI R&D process at Anthropic.ā Fable 5.2, based on its AECI, likely represents āModel 3,ā which is the second-most-capable internally deployed Anthropic model.
Idk man that seems like that pretty solidly means AC did not happen
Relevant interview quotes from the METR report include: Being āno longer bottlenecked on code,ā āSWE acceleration⦠is extremely substantial.ā āWhen people present experiments, they most often just point to an [agent scaffold] thread. Experimentation overhead is driven towards zero super fast.ā Thereās also the reportās assessments on automating researchers beyond coding:
āWe were told that R&D is accelerated primarily via AI assistance rather than fully autonomous contributions, but there are some signs that fully autonomous contributions will play a larger role soonā. āSimply put, Anthropic tries hard to substitute researchers out for models, but researchers are still required.ā Note that humans could still contribute some alongside an AC: the criterion is that only the AIs could do all the coding as well as only the humans.
ok and?? if only the AIs could do the work then they wouldnāt be having trouble substituting out researchers for AI.
In the reportās chart showing āModel 4 achieves >50% on all of its benchmarksā, there is a software engineering benchmark on which Model 4 gets ~52%, and another around ~75%. Surprisingly, on the 52% one, scores have been getting worse since Mythos 5. This makes me discount it substantially, but still itās informative that Anthropic chose to include it in the AECI. The progression for the 75% one looks more reasonable; Eli thinks these benchmarks provide weak evidence against AC. (Not strong evidence, because humans might also score well under 100%.)
My take is that if there was an AI model that could automate all the coding on Anthropicās AI projects, it would be absolutely plastered all over the news. Especially with their IPO looking shakier and shakier by the day. Plus if theyāre apparently trying to replace researchers but canāt thatās a pretty strong indicator that AC isnāt here



Now that would be funny