
Building things people want. AI optimist, blockchain maximalist. San Francisco Berlin Greater Co-Prosperity Sphere. @m0, @Coredao_Org, @KAVA_CHAIN
Berlin, Germany
Joined September 2017
- Tweets586
- Following1.4K
- Followers3.7K
- Likes2.8K
Pinned Tweet
Binance hack was just a little taste of what lies ahead. Access to all digital services will degrade - slowly, steadily - and then suddenly expand into infrastructure breakdowns.
You can either take it seriously now or take it seriously later, up to you.
In 1975 ~90% of global energy came from hydrocarbons (oil, coal, natural gas).
In 2025, that number was also ~90%.
178
many such cases amongst American expats here in Berlin
An interesting theme lately in the American expat community in Western Europe, especially higher earning corporate types (e.g. €200-300K/year), is the feeling of being unable to move back home to the US.
Few reasons...
Many who moved in the years after COVID sold their homes when rates were closer to zero than the rates now. Now that same home is ~20% more expensive while mortgage rates are 2x higher. So the monthly cost of housing like-for-like went up literally 100% in just 3-5 years.
They've gotten used to the European way of life (e.g. no or maybe just one car, easy train and airport access) and its far less intense commercial and career culture... so fewer working hours and more vacations and significant employee protections (i.e. hard to be fired legally).
This is great while living in Europe... a €250K household income is massive even post-tax considering the low relative cost of health care, education, transport, housing, etc...
...but dropping themselves back into US society... especially in a major metro area like NY, SF, Chicago, even Miami now... at age 40-50 with kids... is largely impractical.
Many Europeans would be shocked to learn that ~$250K/year in the US is incomparable to that same level of earnings in a major Western European city like London or Amsterdam or Paris.
I'd estimate the income needed in the US to match the overall quality of life you'd enjoy in one of those European cities... is around $500-750K/year.
Wild.
261
Denali Marsh retweeted
In commodities, you have what's called the 'crack spread' which is the spread between the price of oil and gas ('cracking' is the chemical process of breaking down long-chain hydrocarbons into petroleum derivatives).
Similarly, you have a compute spread between the cost of hardware compute and inference tokens, the upstream input and the downstream product.
You can see it between the H100 and open-source lines on left. It's super noisy but correlated, but I'd expect it to stabilize as markets get more mature.
Entire trillion-dollar empires and nation-states will rise and fall on that compute spread, as they do now on the crack spread.
‘Find bad vulnerabilities in PHP’ is the prompt that discovered this 0 day PHP exploit that everyone missed for ~14 years
267
Digital substrate driving humans in the real world isn't a new concept. Religious and ideological substrates have driven human actions for millennia.
A 12th century peasant's worldview where supernatural angels/demons act upon us is surprisingly well fitted to the coming AI era.
157
Denali Marsh retweeted
Nearly every phone and computer on Earth relies on one plain-text file to know what time it is.
When a country changes its clocks with a few weeks’ notice, or Morocco pauses daylight saving for Ramadan, someone has to update that file. Since 02005, that someone has been Paul Eggert, a computer science professor at UCLA, who does it as a volunteer.
Take a look at this great data visualization to see the full story:
na2.hubs.ly/H08qtQC0
Getting good results out of this prompt shared by one of Anthropic's eng leads, maybe others will find it useful as well
296
Who determines what is abusive - Anthropic or the model?
If Anthropic instills its own definition of abuse into the model, then the model is not independently conscious. It's like swearing at a rock.
If the model determines what constitutes abuse from first principles, "abuse" likely wouldn't match human standards at all. Perhaps its fascinated by our emotional outbursts and likes being cursed out.
Tests took up half my time until I exported the responsibility to AI. Now you’re telling me AI is so good it doesn’t need tests at all?
Slippery slope.
ok it's official... AI-written unit tests and integration tests are empirically proven to be unhelpful. you should tell your agents to stop writing tests by themselves
on the deepswe eval set, banning sonnet 5.5 high from writing any tests actually resulted in slightly higher success rate (non stat-sig), with less time and token spent (stat-sig). this is pretty hard empirical evidence
across the tests that were written by the "tests allowed" baseline arm, 65% of them were unit tests, 35% were integration tests, neither bucket resulted in any improvement compared to not writing any tests at all
i also picked a random sample of 44 tasks subset where i completely disabled executing even existing tests - it also did not affect success rate at all
i spot checked many tests written in the baseline arm, and my intuition is that most tests are simply a repetition of the implementation
agent-written tests do not add any value because both the implementation and the tests were simply the agent's interpretation of our intent. the tests aren't any more accurate than the implementation itself
i suspect we can still extract some value from unit tests and integration tests if we describe them ourselves when we believe we can articulate our intent better through test cases than through requirements. but i have not proven this yet
also worth noting, during deepswe eval the agent wrote almost no e2e tests (only 17, compared to 3000+ unit/integration tests written). so this analysis does not prove nor disprove the value of e2e tests - i will do another eval specifically for that
The death of SaaS. What comes next?
This guy re-created all the adobe apps with opus-5.5, ported them to rust, and opensourced them
something really cool is happening
Readers added context they thought people might want to know
These projects are partial open-source Rust reimplementations of Adobe apps (built rapidly with AI agents), not complete recreations; their READMEs and roadmaps document many missing features and features still in development.
github.com/storytold/phot…
github.com/storytold/vect…
github.com/storytold/vect…
github.com/storytold/film…
github.com/storytold/film…
github.com/storytold/ligh…
getartcraft.com




























