releasing GoatRemote for early access today ⦿ use your mac as smart TV ⦿ vibecoding? we're now couchcoding ⦿ voice commands ⭢ transcription OR actions ⦿ mic is on the remote (or your phone!) ⦿ trackpad + ring scroll for mouse controls ⦿ all running locally link below ↓

was gonna switch from whisper to parakeet for voice transcription since it's faster, but realized parakeet doesn't support text prompt like whisper does to provide context for better accuracy in domain-specific voice commands. guess i'll skip

They built what they thought people wanted. You can build what people actually need. Use Articos and get user insights in 30 minutes before you waste 6 months building the wrong thing. Link in comment 👇

is there a real @Extropic_AI demo showing inference of an energy-based model of matching performance using 10,000x less energy? they mentioned they got it working but i didn't see a demo

if it's just a button to record voice then it's not that interesting

is this like a wiimote in ring form? if so this is my competition for GoatRemote if that's a click button this would genuinely be a very good input device tho i still think a Siri Remote with proper voice commands might be better due to more buttons & trackpad

Always curious and full of energy, Beagles turn every walk into an adventure 🐾🎉.

i made goatremote mac app entirely without xcode and words can't describe the bliss of having all compile, codesign, entitlements, notarize, create distribution app bundle, prepare auto updates, set versions, publish release steps automated in cli scripts

has anyone had luck buying shares in private companies from Hiive? trying to hodl Tenstorrent and Cerebras

recorded music is an anomaly of the last 100 years. it was never really about personal sensory experience before that whatever ai music is ruining for you wasn't lindy to begin with the recordings were never sacred. playing with other people was kinda the point of it all

boring take because whether they are or not leads down two completely different self-coherent frameworks that are so deep it's the most interesting philosophical bifurcation we've ever come across. and the more you explore the more you'll have opinions i guarantee

dbpill (the Postgres proxy that automates index optimization) is now open source: https://t.co/8eUivZTNUP enjoy

on a separate note i think another bizarre point about consciousness is that it can only ever be talked about when there's memory of it the completely *in the moment* version of conscious experience has no way of communicating with the outside world without memory & speech, and

one weakness of computationalist theories of consciousness is that they fail to explain why a brain might have separation of conscious and unconscious thought when all the information is integrated anyways it fails to propose where borders might lie, whereas physicalist theories

docs here https://t.co/40VrnlZSM9 and example here: Vid: https://t.co/DNXCGRCXrt Transcript:

this turned out to be accurate the first good diarization seems to be `gpt-4o-transcribe-diarize` trying it right now, it can handle most situations of people speaking over each other & label them correctly. still not perfect obvi. and expensive, $0.30 for 10mins

here are a few other tips: 1. use completion mode, but include chat tokens <|im_start|> etc, and prefill response start so LLM follows structure perfectly 2. make sure changing parts of the prompt are at the end only, to make use of cache well 3. parse output stream manually

this was the joy sparker btw. finally a local use case that's reliable, fast and useful https://t.co/8JWS7mHp1Y

here's the GoatRemote prompt for turning voice commands into macOS actions (transcription, keyboard shortcuts or applescript) model: qwen3 instruct 4b @ q4 it's a 5000+ token prompt, but keeping kv cache in mem means response done in 0.25s! https://t.co/IfQTo73EsI

@zaygranet come on bruh do the right thing you're not saving Tim Apple but you're killing me

the testimonials are coming in

🔥 Today we’re excited to announce new funding for `grep` (at a $1.3B valuation) to continue building the foundation of agent observability and text search infrastructure. grep began as a humble UNIX utility in 1973. Since then, it’s evolved—through recursive innovation and the

might just have product market fit for https://t.co/oduHU3kEm9, not bad for one tweet so what's the play for producthunt these days?