henry 🌘

@1thousandfaces_ reading my friends posts after a long day on the frontier

@aisconnolly job-boards.greenhouse.io/anthropic/jobs…

New hobby: reading privacy data engineering job ads and daydreaming about what it would take to replace the entire personal-data stack with ZK / MPC so the company never has to store the data in the first place. explore.jobs.netflix.net/careers/job?do…

@inerati yes, just like process isolation, it’s possible to end up in a logically corrupted state. in practice (ime) lower level corruption than that is very very rare except possibly when doing bad things with native libs

@hdevalence hmm couldn’t this leave things in a corrupted / bad state tho?

i’m very happy with tokio

btw, if you want to use some nice ideas for “functional programming for network services” that are slightly more recent than Erlang, study the tower::Service trait and “your server as a function”

@inerati unfortunately claude loves to “defensively” avoid crash propagation without thinking about what this means for fault isolation / recovery. i assume this will get better over time

@inerati yeah, panics in tasks kill the task; if you retained the join handle to wait on the task, you get a joinerror, so you can either .unwrap() to propagate the crash or you can handle it somehow, depending on where you want to draw fault isolation lines

@inerati cancellation can be annoying but crashing tasks works fine ime?

@hdevalence i wish you could actually let it crash or cancel in a sane way tho

m.soundcloud.com/aboveandbeyond… one of the best

bc there's a crack in everything and it's much easier to not leak if you have a culture of not flirting with the line

@amphichrome_ it kind of is. photo: carbon deposition on silicon substrate

what if the earth was a wafer

My CENTCOM told me the IRGC keep bombing his THAAD batteries so l asked how many batteries he has and he said he just goes to Lockheed Martin and gets a new THAAD afterwards so I said it sounds like he's just feeding THAADs to the IRGC and then INDOPACOM started crying.

@tenobrus uh, yes?

"don't worry about the impact of ai on cybersecurity guys, agents will help us harden all that too, defense is more powerful than offense"

@SurrealistShip @Duderichy nope, guess i can add spotting impaired credit to spotting tigers as activities for which i am genetically disadvantaged

@hdevalence @Duderichy Does inverting colors work for colorblindness?

bruh

@SurrealistShip @Duderichy i’m colorblind

@Duderichy Spot the outlier

@tenobrus assortative processes rule everything around me

@tenobrus selection effects

sometimes i take a look around and notice its pretty weird that all my friends seem to have like 10-50k followers on twitter. sort of unusual position to find urself in anthropically speaking. what's the deal w that ?

@fede_intern @Zac_Aztec @ClementWalter @zkGaylord @diego_aligned @VitalikButerin @alexhooketh @GeoffreyHuntley @eth_proofs these seem like tractable problems to me. ymmv

yes, a with everything the AI can do everything correct. however it can't yet, it does too many subtle mistake. 1. the training dataset is small 2. the consequence of an error are huge 3. there is no easy feedback loop for it to learn with time i think it will do it better than any human.

This is quite an impressive experiment. Vibe-coding the entire 2030 roadmap within weeks. Obviously such a thing built in two weeks without even having the EIPs has massive caveats: almost certainly lots of critical bugs, and probably in some cases "stub" versions of a thing where the AI did not even try making the full version. But six months ago, even this was far outside the realm of possibility, and what matters is where the trend is going. AI is massively accelerating coding (yesterday, I tried agentic-coding an equivalent of my blog software, and finished within an hour, and that was using gpt-oss:20b running on my laptop (!!!!), kimi-2.5 would have probably just one-shotted it). But probably, the right way to use it, is to take half the gains from AI in speed, and half the gains in security: generate more test-cases, formally verify everything, make more multi-implementations of things. A collaborator of the @leanethereum effort managed to AI-code a machine-verifiable proof of one of the most complex theorems that STARKs rely on for security. A core tenet of @leanethereum is to formally verify everything, and AI is greatly accelerating our ability to do that. Aside from formal verification, simply being able to generate a much larger body of test cases is also important. Do not assume that you'll be able to put in a single prompt and get a highly-secure version out anytime soon; there WILL be lots of wrestling with bugs and inconsistencies between implementations. But even that wrestling can happen 5x faster and 10x more thoroughly. People should be open to the possibility (not certainty! possibility) that the Ethereum roadmap will finish much faster than people expect, at a much higher standard of security than people expect. On the security side, I personally am excited about the possibility that bug-free code, long considered an idealistic delusion, will finally become first possible and then a basic expectation. If we care about trustlessness, this is a necessary piece of the puzzle. Total security is impossible because ultimately total security means exact correspondence between lines of code and contents of your mind, which is many terabytes (see firefly.social/post/x/2025653… ). But there are many specific cases, where specific security claims can be made and verified, that cut out >99% of the negative consequences that might come from the code being broken.

@1thousandfaces_ fact check: