Dan

→ Airbnb for LLMs. Run or rent open-source models like LLaMA & Qwen. Building in public. Waitlist open:

[Dan @comcava]

Wanted to watch Terence Tao's lectures in Russian in his voice. That audio track didn't exist, so I built it: a service on NVIDIA DGX Spark that turns subtitles into voiceover. Upload a few voice samples,voilà. Batch uploads supported. Live for everyone🎙️ https://t.co/fDIvt5PQ3j

[Dan @comcava]

@AlexFinn I'm curious what you use for orchestration. I used «opencode» and ended up with a non-functional application that crashed, after several iterations of fixing it in the paid model.

[Dan @comcava]

@iam_smx The Moon is basically a giant radiation grill with no magnetic shield. No topological qubits -no Moon computers.

[Dan @comcava]

@danieldaneshi @yishan Moats don’t exist in GPT land , only momentum.

[Dan @comcava]

@johnrushx @yishan https://t.co/JIj8lGifLW?

[Dan @comcava]

@danieldaneshi @yishan Nah, he’s just saying the funding isn’t enough to build a real company - not that infra guys would crush you. Others could just copy it easily.

[Dan @comcava]

New grant idea: a Svalbard‑for‑Synapses vault—terabytes of human know‑how sealed in a lunar lava tube. That way, if we self‑delete, the first AI archaeologists boot up with Wikipedia and cat memes pre‑installed. Who’s backing “Project Doomsday Dropbox”? #BackupTheSpecies

[Dan @comcava]

Notes from the (probably) robotic future…

When the first hazard‑proof workbots roll out, humanity will finally outsource nuclear meltdowns. “Chernobyl 2: Electric Boogaloo” will trend while the robots calmly colonize the exclusion zone and call it home. #FutureFriday #Oops

[Dan @comcava]

Traffic update: autonomous cars now zip along a new “silicon‑only” hyper‑lane—top speed: “LOL, you blinked.” Flesh‑and‑blood drivers are quarantined to the slow‑moving protein lane, where turn signals remain an unsolved mystery. #RoadsOfTomorrow

[Dan @comcava]

Can we stop calling it “vibe coding” and start calling it “coding‑by‑touch (a.k.a. blind coding)”? Fixing typos at today’s prices feels like paying yacht money for a rubber duck. #DevLife

[Dan @comcava]

5/
We’re just getting started — and opening up our waitlist.
If you run models, or want to earn from your idle GPU:
→ https://t.co/Ttb5kvrXZO

Let’s make LLM inference accessible.
Simple. ⚡

[Dan @comcava]

1/
One day I realized:
Our e-commerce project needed tons of inference — for matching competitor prices using image & text embeddings.
We didn’t need GPT-4.
We needed scalable, fast, open-source models.
And… they worked. Perfectly.

[Dan @comcava]

4/
Atlora is a marketplace for LLM inference:
→ Need to run a model? We’ll match you with GPU providers.
→ Got a GPU? Get paid to host open-source models.