jeremie.com
Dream things. Make things. š«¶all. He/him.
(wayback created Jabber/XMPP)
175 posts
1,700 followers
152 following
Regular Contributor
Active Commenter
comment in response to
post
I would love to hear about your research here! ā¦ and block all those unthoughtfully dogpiling
comment in response to
post
This is like, šÆ Black Mirror S07E01, yet another episode that Iāll never forget š
comment in response to
post
Iāve been wondering on using public-key encryption for this, where the blobs (perhaps even full CARs) are encrypted to a public key and the private key is distributed OOB (DMs, MLS, etc).
Allows for them to be openly published and attached to social events, but only certain recipients can decrypt.
comment in response to
post
Agree completely on the "meh" smell and the hype dogpiling karma farming rapid exploitation of MCP. It was launched with a very niche specific use-case and is now being bikeshedded to all get-out...
That being said, it'll have a place and provide some value, I just don't find it very interesting š¤·
comment in response to
post
Pugnocrat? Pugnacious + Autocrat
comment in response to
post
Thank you for publishing this along with the talk, itās a wonderful historical reference for the founding tenants of #Atproto!
Kudos to you and the rest of the team on intuiting your way to these principles and persistently fostering it into a thriving service š
comment in response to
post
@atprotocol.dev will you be posting the recordings of individual talks? Blaneās would be great to drop here š
comment in response to
post
comment in response to
post
I donāt think itād be that difficult, the did:plc directory could add support for an alsoKnownAs with a different URI prefix + some logic to verify it w/ a clear deconfliction policy: github.com/did-method-p... cc @bnewbold.net
comment in response to
post
comment in response to
post
Just at a glance this seems to require an already compromised firmware or exploit on the esp32, the ābackdoorā commands they describe are instructions sent internal to the chip.
comment in response to
post
Orville easily in the top 5 š
comment in response to
post
I saw and am highly displeased. Where is the airlock...
comment in response to
post
I really hope Apple News can add something like Connections
comment in response to
post
I donāt know what a āsmartstrapā is but having some I/O ports and a wee bit of power available (out, and maybe in too?) via the strap attachment could make for a very interesting extension mechanism ;)
comment in response to
post
Devs was amazing
comment in response to
post
"Local macro enthusiast dismayed at widespread informal colloquialisms, full report at 10."
comment in response to
post
Itās everyoneās desire for this to not be a ācompany-controlled platformā, already built on an open protocol, but as you know fully federated infrastructure is more than just documenting the bytes on the wire ;)
comment in response to
post
IMO it became diluted to crap years ago already by human regurgitated content on top of human regurgitated content.
comment in response to
post
Gave it 25 loops, it certainly embellished to an extreme, but it's clearly (IMO) not better than the original, you're still safe for now š
github.com/quartzjer/lo...
comment in response to
post
Did you read Perfect Mirror already? š github.com/quartzjer/lo...
comment in response to
post
Feeling a bit better today, I added some improvements to loopy and a couple more stories:
* switched to markdown (light formatting and easy reading on github)
* add git support to track each loop revision
* ability to stop and override the editor feedback for a run
* easy bootstrap story from stdin
comment in response to
post
Actually, the little project I posted yesterday is a decent example, feed in a basic short story or just some ideas for one and it embellishes to an extreme by sampling the probabilities around the inputs: bsky.app/profile/jere...
comment in response to
post
Once I realized that their training created clusters around concepts, and that the prompt input was allowing you to direct the sampling of statistical probabilities around and between those clusters, the results started to feel a lot more intuitive.
comment in response to
post
Itās the easiest minimal client, excellent on-ramp into the API and dev ecosystem: github.com/quartzjer/AT...
comment in response to
post
I've also incidentally now become a maintainer on xclip š¤¦ github.com/astrand/xclip
comment in response to
post
His fav spot this time of year š»