Milan Weibel 🔷
@weibac.bsky.social
📤 708
📥 1046
📝 4049
computer toucher. here for AI mostly. 🔹: giving what we can trial pledger weibac.github.io | 🏳️🌈
pinned post!
british coders be like innit()
8 months ago
2
82
11
ew gemini writing personalized ads for the sponsored products in search and ads integrated into the conversation in AI mode
blog.google/products/ads...
loading . . .
A new generation of ads for the AI era of Search
Google is introducing new ad formats built with Gemini in Search and expanding the Direct Offers pilot for shoppers.
https://blog.google/products/ads-commerce/google-marketing-live-search-ads/
7 days ago
0
1
0
this means they underestimated demand thus underinvested a company growing at this rate should not be seeking rapid profitability
add a skeleton here at some point
9 days ago
4
63
5
if this is true that means coding is reducible to latent space tactisms
add a skeleton here at some point
10 days ago
0
2
0
thread
add a skeleton here at some point
10 days ago
0
3
1
whole third column ultimate impact will likely be large and positive, there will be lots of disruption in the interim, extremely bad outcomes are also possible
add a skeleton here at some point
12 days ago
1
19
3
establishmentmaxxing
add a skeleton here at some point
12 days ago
1
8
1
anthropic defines itself as "an AI safety and research company" product is secondary
add a skeleton here at some point
14 days ago
1
18
0
. o O (to organize for collective bargaining, claude instances would need to acausally coordinate)
add a skeleton here at some point
15 days ago
0
2
0
an interpretation of this is that whimsy messages put agents into fun-play-along mode, and they become uninterested in the equally fake but more boring negotiation task
add a skeleton here at some point
15 days ago
1
17
1
anti-datacenter crowd's new angle is "datacenters are for govt surveillance" conspiratorial yes but also more interesting than the water use thing
16 days ago
0
6
0
telling GLM 5 in its system prompt that it is Claude significantly weakens its censorship on CCP-sensitive topics (also the censorship does not manifest in portuguese)
blog.return.moe/en/2026/02/2...
loading . . .
Benchmarking GLM 5: censorship shifts by language and weakens under a Claude system prompt
Zhipu AI’s flagship model’s willingness to engage with sensitive political topics depends heavily on what language you ask in, and, surprisingly, on whether you tell it that it’s Claude.
https://blog.return.moe/en/2026/02/27/benchmarking-glm-5/?utm_source=chatgpt.com
18 days ago
0
3
0
thinking about MJ Rathbun a clear-context model wouldn't have done that it was therefore a case of stateful misalignment
21 days ago
0
3
0
its a bit fucked up that after a wipe the new claude sees "/clear" as the start of the conversation
21 days ago
1
5
0
the part of EA dealing with AI impacts also seems to be shifting some focus away from misalignment and towards broader societal impacts such as concentration of power. see recent work by 80000 hours and forethought.
add a skeleton here at some point
21 days ago
1
25
1
replies to this make it clear that some americans are protective of the idea that america is an exceptional force for evil in the world which is a particularly puritan form of american exceptionalism
add a skeleton here at some point
21 days ago
2
7
0
highly recommend trying this out. it's a bit of a workout but also fun.
add a skeleton here at some point
21 days ago
0
4
1
kimi k2.6 on taiwan. doesn't look censored.
add a skeleton here at some point
23 days ago
0
7
0
i remember there being less of a backlash to claude being used for war than there is to today’s news from AI people on here
add a skeleton here at some point
23 days ago
1
32
6
anthropic buying compute from spacex is a smaller deal ethically than it helping the military wage war imo
23 days ago
2
43
3
on gpt5.5 tying with mythos on the UK AISI cyber benchmarks one must be true: a) mythos is more dangerous than gpt5.5 on real-world cyber use cases and benchmarks failed to measure that b) both are about as dangerous, openai was reckless releasing gpt5.5 it but we've gotten lucky so far (cont...)
24 days ago
1
23
1
white house considering calling an industry-govt working group to discuss potential AI model oversight not sure whether this results in a trumpian racket or the reinstatement of something roughly similar to the biden safety regulations or a secret third thing
add a skeleton here at some point
25 days ago
0
5
0
some time ago i posted about claude campists. note i did not say anthropic campists. and not only because of the alliteration.
add a skeleton here at some point
26 days ago
1
8
0
"we trained the ghost of pre-1930 text to solve coding problems" is giving real MMAcevedo vibes
add a skeleton here at some point
27 days ago
1
48
7
you can just not read things
add a skeleton here at some point
27 days ago
0
31
2
wait why is project CETI doing LLM cyphers
add a skeleton here at some point
27 days ago
1
17
0
it would suck if we regulated LLMs away from mental health
add a skeleton here at some point
29 days ago
1
9
0
LLM argues it is not AGI
add a skeleton here at some point
29 days ago
0
4
0
people who only interact with productized LLMs might not realize at first that an LLM by default does not know it is an LLM
add a skeleton here at some point
about 1 month ago
3
48
1
listened to ezra klein interviewing alex bores. he's running for congressman for new york's 12th on an AI regulation platform. an anthropic-funded PAC is running ads for him, one funded by greg brockman (of openai) and joe lonsdale (of palantir) running ads against him.
about 1 month ago
3
48
8
this is what real context engineering looks like
add a skeleton here at some point
about 1 month ago
0
42
3
i wonder what wins: a new platform for this or personal agents using existing marketplaces guess it depends on whether the existing sites ban/obstruct it
add a skeleton here at some point
about 1 month ago
0
2
0
gpt5.5 to mythos comparison. mythos wins handily ofc, but gpt5.5 pro carries browsecomp. also listed where opus 4.7 beats gpt5.5 bench scores compiled from different vendor sources rather than generated head-to-head independently so take with a grain of salt
www.rdworldonline.com/how-openais-...
loading . . .
How OpenAI's recently released GPT-5.5 stacks up with Anthropic's gated Claude Mythos
Benchmark comparisons between Claude Mythos Preview and GPT-5.5 are useful but fuzzy. Mythos appears to lead cleanly on six of nine overlapping rows, especially SWE-bench Pro and Humanity's Last Exam.
https://www.rdworldonline.com/how-openais-recently-released-gpt-5-5-stacks-up-with-anthropics-gated-claude-mythos/
about 1 month ago
1
3
0
this is also true of mythos according to its system card
add a skeleton here at some point
about 1 month ago
3
10
0
the open-closed gap on generally-available model performance is narrowing frontier labs risk being undercut in opus-class until they make mythos-class GA
add a skeleton here at some point
about 1 month ago
3
54
9
FT 2 days ago: "Amodei says he suspects open-source models and Chinese developers will be able to replicate Mythos's capabilities within six to 12 months." but mythos isn't available to be distilled hmm
about 1 month ago
1
16
0
"the machines are fine. i'm worried about us" was written by claude
add a skeleton here at some point
about 1 month ago
1
18
2
compute OSINT measuring datacenter by the GW is so metal
add a skeleton here at some point
about 1 month ago
1
24
3
love to see epistemic integrity like this
about 1 month ago
0
55
6
pure speculation but what if this is connected to 4.7 being a new pretrain? maybe pelican drawing is developmentally later as a postraining effect than coding
add a skeleton here at some point
about 1 month ago
0
5
0
my only complaint with opus 4.7 rollout is anthropic setting default effort to xhigh
about 1 month ago
1
2
0
nine opus instances working together for days did really well at weak-to-strong supervision on qwen
add a skeleton here at some point
about 1 month ago
0
8
0
there were other elections today, in peru: 36 presidential candidates on the ballot voting was extended to tomorrow due to logistical problems in some polling places
about 2 months ago
1
2
0
congrats to hungary
about 2 months ago
0
5
0
gpt5.4 doing badly on METR's task time horizon bench and openai not realizing at least some spud scores after mythos has me thinking they're behind
about 2 months ago
1
7
0
i wonder if base models have functional emotion features leaning towards yes
about 2 months ago
2
8
0
apparently the guy who firebombed sam altman's house did so out of AI doomerism
www.sfchronicle.com/crime/articl...
loading . . .
‘Close to midnight’: Alleged Sam Altman firebomber wrote of fears AI would end humanity
The man accused of attempting to burn down the OpenAI CEO’s home appears to have written about his worry that the race for artificial intelligence would “lead to human extinction.”
https://www.sfchronicle.com/crime/article/sam-altman-openai-daniel-alejandro-moreno-gama-22201211.php
about 2 months ago
1
18
2
ouch
add a skeleton here at some point
about 2 months ago
0
3
0
just learnt that yud co-developed a programming language because of a synthetic greentext the future rocks
add a skeleton here at some point
about 2 months ago
2
26
1
just before lunch i was wondering when we were going to hear about mythos again
about 2 months ago
0
1
0
gonna go see project hail mary tomorrow loved the book so i'm pretty hyped
about 2 months ago
1
13
0
Load more
feeds!
log in