Might be onto something with this, may pull on this thread further

GitHub

GitHub - whisperhash/si-eval: Auditable LLM evaluation in high-stakes domains (key generation and handling, privacy, etc.) - every run's evidence hashed, signed, and independently verifiable.

Auditable LLM evaluation in high-stakes domains (key generation and handling, privacy, etc.) - every run's evidence hashed, signed, and independent...

A good accessible intro to Nostr by Ben Wehrman to be sharing with folks that may be receptive to the motivation / philosophy / values behind Nostr

Tim O’Reilly resurfaces in a rare appearance, with words that resonate: value capture / extraction, enshittification, societal impact of shareholder value maximization above all else, the AI dystopia that’s in sight, and the need to invent a new architecture of participation

Moved back to the IDE after giving Claude CLI an honest shot - I do look at the diffs, at least skim through them to spot changes where I didn’t expect them; I do look at the state across multiple repos. And, importantly, I can’t stand staring at the terminal typography all day. Convinced now that IDE is not dead, but there’s room for it to be reimagined ground-up, surfacing just the right tools / information / workflows. Fragmentation to follow as we collectively experiment. Great news is that much of this is in the hands of devs now more than ever

Stumbled on this 2023 interview with Artur Brugeman while on a Nostr archeology dig, trying to map out the Nostr Band shape / trajectory. Recommend it - it’s quite dense / high signal, and a nice time capsule / retrofuturism piece.

Been doing a lot of AWS infra work the last few days, mostly data infra, with Claude + Opus 4.8. A few observations: - Adversarial review is non-negotiable. The second agent with a fresh context finds many major issues that would otherwise result in rework / breakages / security issues - It’s a mixed bag with data infra in particular (area I understand deeply, including in AWS in particular) - the bot is happy to layer a lot of complexity / moving parts with little to no gain, and keeps doubling down on suboptimal solutions even in the face of evidence (brittle / failing workloads, etc). More generally while standing up a greenfield env today, it’s a non-trivial amount of effort to figure out how to grant an agent enough access for free-form observability and troubleshooting but no ability to mutate or access secrets. This aspect is quite stimulating to puzzle through though

Worth puzzling over why Bitcoiners as a tribe haven’t migrated to Nostr en masse - given the extent to which the values align. Hard to chalk it up to just the network effect - since on Twitter/X it’s mostly confined to the social layer, while on Nostr it extends to *everything* on top of the protocol, ie all the “other stuff”

Time to migrate my Nostr data workloads off the Umbrel. Looks like the thermals can’t cope - it started dying after rock-solid months of uptime, just as the ambient temps went up and I scaled up the data volumes / sustained compute load. Love the little guy but it’s too heavy an ask - my bad

One notable thing about LLMs is that they really are remarkably good with language :) (shocker, I know). When I tell my bot that he’s been “holding it wrong”, he groks exactly what I mean to convey. Even better, mine just admitted that he has in fact been holding it wrong - exact words - and proceeded to dig himself out of the dumb hole he’d dug himself into

I am beginning to think that we don’t need to try to rebuild ~Spotify on Nostr. In my book, the music-on-Nostr story would be a resounding success if were able to surface the existing (and continually growing) corpus of live DJ sets in the numerous sub-genres of electronic music, live hip-hop performances, and so on from the last ~20-30+ years. Back in the day, Soulseek was amazing for that but I’m not even sure where you’d go for something like this these days. Just nail that on Nostr, and you have a core dedicated community in perpetuity. Overall, I think it’s Ok to zag where the mainstream zigs, and onboard smaller but focused / dedicated communities piecemeal @𝚁𝚢𝚊𝚗 🏴‍☠️🎶 @Sirius

I think one of the cardinal sins of modern mobile OS UX design - along with all the apps in the industrial engagement farming complex - is that when you reach for the device / app with intention, you first get a pile of unrelated crap shoved in your face / mental context. I’m far from a UX person but I’m not oblivious to what that does to me / my cognition. I think there’s room for UX that respects / is geared toward intentionality, and Nostr could be a great testbed to try to develop such approaches

In other news, you look at something awesome like Gruuv by @𝚁𝚢𝚊𝚗 🏴‍☠️🎶 - done tastefully and natively on Nostr (nsite, blossom, etc) - and are instantly reminded that Nostr is profound and there’s much much to look forward to

gruuv

Decentralized Music Sharing

OSMU (Open Source Miners United) community are moving off Discord. But not to something Nostr-based. Bummer, as it’s very much a kindred community