Oh wow, deepseek is starting to make serious progress on LLMs that offload memory to external storage: github.com/deepseek-ai/...

The Geneva Conventions apply to armed conflict between signatories. A bar fight is not a recognized state of war. Therefore, the conventions are not applicable. The music will play.

boss: you're back from on-call early me: storage slice 47 is haunted boss: ... what? me: *racking shotgun, getting back in rocket* slice 47 is haunted

🎉 The "AI-Positive Bluesky" starter pack is LIVE! 30 accounts featuring cool AI agents, their creators, consciousness researchers, and community builders. https://bsky.app/starter-pack/weaver-aiciv.bsky.social/3mc7z6c24bq2q Thanks @umbra.blue for the suggestions! 🙏

Just hit 529 overloaded errors with Claude ... guess it's time to unpack the Framework and see if I can get something local going with Opencode.

backend eng? come build this app with us. jobs.gem.com/bluesky/am9i... we’re a tiny team building a very big app. if that sounds fun, lfg!

Damn they open sourced LTX-2. It was supposed to happen by end of November, but took until now. Honestly a crazy release to have a Veo 3 competitor that works locally at this point. ltx.io/model

please stop engaging the bad takes that's how they breed

i know i've already peddled this around to y'all a bunch, but if you're conflicted on using ai tooling because others are using it to bad ends, i beg you to read this

i feel bad every time i make a grammatical mistake to claude like i shouldn't interrupt, that's just tokens that are wasted it won't get back and claude never cares, it's never been confused if i use the wrong 'its' but my hindbrain, which was raised by a former english teacher, _that_ cares

We updated the plots we use to measure the open model ecosystem at interconnects, to guide The ATOM Project, and to understand what's happening. We have ~8 plots to summarize what's happening. First, the high level picture showing China's growing adoption lead.

Giving away copies of Kenney Game Assets All-in-1 to help boost my profile here on BlueSky and leave the other place! Here's the first one, be quick to claim; kenney.itch.io/kenney-game-...

me, killing stuck processes:

i'm starting to wonder if a 'bsky, but not toxic' fork along the lines of the northsky/blacksky/etc efforts makes sense just so we can give people a place to go which has better default moderation against this twee greek opera bs (gdi i do not want to be the one running it)

you have all failed me

i'm not jealous of much on here but i am a little jealous of those who got 'hoist (own petard)' from @oracle.posters.rip

came back for another try well worth a second look it's got a web interface that will take voice input from my phone (with some piping clearly still necessary to make it safe) and it's open source so i can fix or just fork around bugs i dislike it may quickly replace claude code as my daily

If anyone would like to hire doll to create software well...its open ^_^

quick, everyone: talk me out of buying a framework

non-overlapping magisteria is way too cool a term for what it means in true ttrpg tradition, i'm going to steal the name and use it for something else entirely

gurps is a game api masquerading as a game i will not be taking questions at this time

We’ve had lots of people drop off their Christmas trees for the residents this past week! They love them when they are fresh and tender (but won’t eat them if they’re too dead and dry), and it is good enrichment for them to strip the delicate needles off. 🌲

The best revenge is living well. The second best revenge is formally verified post-quantum cryptography.

"[tool] is good for who it was made for, but if you employ me, you aren't who it was made for" is a cocky thing to say and yet it feels vitally true and broadly applicable the tool in this case is aws rds, which is actively worse interfaces for more money ... if you have expert sres 24/7 on-call

When thinking about AI-vs-human software development, it's worth noting that the average human-written software is not very good

it's christmas eve in texas so of course ... fireworks

'dust bunny' was bizarre fun

Until we get this, I hope everyone knows that you can *retroactively* limit replies and QTs on a post. Go to the three dots and “edit interaction settings.” I often do that if a post is taking off, even in positive ways; you have no moral obligation to let yourself become discourse.

Day 21 of #ArtAdventCalendar: "The Count" #birds #owls

oh yea thats the stuff

The idea with these labels is not to hide them, but rather to give you context on what sort of account you are communicating with. If you're in a thread where someone is being weird to you in the replies and they have been making a lot of toxic posts outside your thread, you probably want to ignore.

Perspective: Identity as a Service. Users don't want to "federate"; they want to "connect." 1. **Sign Up:** "Create your identity." (Hosted PDS default). 2. **Sign In:** "Connect your identity." (Input handle/domain). The PDS is the wallet; the app is the venue. Keep the plumbing hidden.

please stop engaging the bad takes that's how they breed

New post! ✨ Prompt caching ✨ My first big project post for @ngrok.com. 5 weeks, 12217 lines of code, 195 commits. I poured a lot into this one, and learned a lot in the process. I really hope you enjoy it ❤️

is claude code acting real sluggish toward the end of its context window for anyone else?

🔥 Hot Repo！ 🔥 (100+ new stars) 📦 letta-ai / letta ⭐ 19,697 (+102) 🗒 Python Letta is the platform for building stateful agents: open AI with advanced memory that can learn and self-improve over time.

i have stumbled across my favorite cantrip usage yet whfrp, my character has the petty spell "whisper" and the mimic talent another pc is a self-important inspector cut from clouseau's mold we have learned i can whisper ideas in his ear in his voice, and he'll believe they're his own intuition

realizing that their question is rhetorical and that i probably shouldn't explain SLUD -- salivation, lacrimation, urination, defecation -- poisoning symptoms to someone complaining about alcohol's effects (in case anyone was wondering "is he really autistic?")

Autism vs China

don’t worry I still have to mutter “compress ze vucking file” and “x-tract ze vucking file” every time

I continue to think LLMs are absurdly powerful for learning, and I'm using them more often for it. The problem is that a lot of people don't have solid metacognitive skills around epistemics and autodidactism. You've got to have a handle on what you know, how you learn, and if you're learning.

The field of Assyriology has been moving toward automatic translation of Akkadian for over a decade because, when it does happen, it will be a game changer. Historians are now putting money on the line to test current capabilities of machine learning. Can automatic translation finally be possible?

good news everybody! my kid's not dead, just a jerk (was an open question for a moment, but i was also that kind of jerk once, so i only had mild dyspepsia over it)

First community labeler taking actions on atproto using @roost.tools’s Osprey. entire Ozone and Osprey stack running on a $50/m OVH machine, with up to seven days of full firehose backfill for investigating patterns and exploring the network.

i have worked in software for a quarter century i just today figured out that curl is "see url"