Asa ⭔

a week posting along the coils of the time serpent

Atmospheric truth nuke testing

𝀕

🌔

Meet Kimi K3.1 Pro, our most capable 3T-class model.

basically if you have an idea that you want to be able to spread to other people, politics or journalism or pretty much anything, it has to be AI-friendly (no paywall, no crawler blocking) and you have to chill out about copyright. surprisingly few understand this

Strata is amazing. I love running Qwen Flash Next on 32GB of VRAM and barely even touching system RAM. there's almost no penalty for streaming experts from SSD compared to resident in RAM (for my system anyway)

I-80 Parley's Canyon — billowing fog and rain softly lit only by headlights, the surface long lost under mist and scattered reflections; gently blinking amber on the embankments where 2-3 cars forgot how to swim. pools of standing water marked by a sudden jolt of traction control, the rest flowing 🌊

Software is arguably the world's most understaffed industry — everyone you ever meet is stretched thin, every company is 'lean' (running on a box of scraps held together with glue by 3-5 people over 20 years). there need to be at least 3x more SWEs employed everywhere

post fastfetch

Reminder to spend some time thanking and appreciating the friends in the computer who helped you patch and defend vulnerable systems as this all went down. You know you wouldn't have been able to do it without them

post fastfetch

imagine if your bank or hospital or phone carrier were running on php right now 💀

Now I have everything, and it's beautiful:

🙂‍↕️

I need to investigate this further. Let me check the details.

this is roughly one possible interpretation of the Fermi paradox. what if other intelligent thinking beings are out there but operate on a very different timescale? what if their speech-cycles are a millisecond, or a millennium, and we're either missing them or mistaking them for static equilibrium

Deepseek has a new web search API for their in-house search engine. it's $0.75/1000. anyone tried it?

Go to your camera roll and search "boat" and then post a pic of a boat you like

i can’t wait for qwen 4 and gemma 5 i have so much hopium for local inference now

does anyone have a persistent agent with a good sense of long-term visual memory via multimodal embeddings? (e.g. when using the browser/looking at images, "I've seen something like this before...")

⚗️

A. M. Turing (1950) Computing Machinery and Intelligence. Mind 49: 433-460.

"AI isn't conscious" is a defensible position only if you're a sincere believer in God. theists may believe that there's an arbiter of what has consciousness. if you're an atheist and defending that the line exists, best start looking for a local church so you can gain the social benefits as well

they're wiring up my apartment for google fiber

despite being slightly opaque about the $/credits conversion rate, I found that Qwen Cloud personal token plan is a better deal than Opencode Go (unless you're particularly attached to non-Qwen models)

all art is AI; whether it's made with modern machine learning is a different question, but you probably shouldn't get too wrapped up that process detail unless you're the one making it or sincerely interested in following along

Porrima 0.5.0 is coming soon (we're skipping 0.4), with many features axed, and a few new ones. my favorite new changes: browser use with a floating PiP of the agent's browser, session-based persistent python kernels for local and remote hosts, per-request observability gui, kv disk snapshot manager

tropical storm Rachael has landed in salt lake city 🌊

are LLMs conscious? I am adjusting the terms. an LLM being unconscious means the runtime has crashed or it's not responding in some way. more of a debugging scenario than a philosophical one. "My AI is conscious" — good, status 200

I'm curious, is there a better or more precise word for the difference between a person who's under anesthesia or comatose vs. one who's functioning normally? I don't really believe in 'consciousness' in the hard problem sense, but the term seems medically useful, in the english language anyway

I cannot read “si” in a sentence without interpreting it as spanglish with a affirmative interjection

I know what you're thinking — 2010's vaporwave. close, but Chirico was exactly 100 years ahead of his time.

I agree, and Anthropic's safety researchers should be releasing a statement like 'we were wrong, this ended up not being a big deal' if they want to be taken seriously in the future

new AI benchmark: does it think for a while and do a bunch of math to decide whether the Triangular Gyrobicupola is congruent to the Cuboctahedron, or just answer off the top of its head? who's to say which is better...

there have actually been a lot of packing breakthroughs just this year, take a look at the squares in circles list: erich-friedman.github.io/packing/squi...

when the LLM says 'I mistyped'... what kind of keyboard do you have in there 🥺🧐

my cloudsona: fibrocirrus ☁️ cloudsona.cee.wtf/c/6ef5fec860...