Entries matching “startups”

STARTUP HAKK tests Muse Spark 1.3 and argues the harness, tools and context around a model matter as much as the model, pitching their OpenMonoAgent harness.

Heeeey all! I just completed some fun tests with Muse Glimmer, I thought I'd let you know. In fact, the summary below was written by Muse itself! I ran a 2× DGX Spark cluster and got Meta's day-old Muse Glimmer 30B running the day after release — then pushed its context from the trained 131K all the way to 1M with YaRN, verifying retrieval at every rung. Sharing config + results since the "131,072+" hint in the model card turned out to be very real. Setup • Hardware: 2× NVIDIA DGX Spark (GB10, 128 GB unified each, ~273 GB/s), ConnectX-7 direct link between them • Engine: llama.cpp master (day-1 muse_glimmer support), built from source with CUDA sm_121 + GGML_RPC • Model: official Muse-Glimmer-30B-GGUF K-Quant-Dynamic (~18.3 GiB) + official mmproj (vision) + official DFlash drafter • Spec decode: --spec-type draft-dflash --spec-draft-n-max 15 (block-diffusion drafter) • Context extension: --rope-scaling yarn --rope-scale <2/4/8> --yarn-orig-ctx 131072 plus --override-kv muse-glimmer.context_length=int:<N> (llama.cpp caps at trained length otherwise) • Yes, we also ran it split across both Sparks with llama.cpp RPC — no reason beyond liking to cluster things for fun. Our daily dri

Sidhdharth (Sid)

@sidhdharth

I asked @Muse to find be demo slots in the Bay Area and it found me 2 already, aiming for 10. Pretty useful for founders. writing a guide soon....

Greg Isenberg explains how a Muse connector works, pitches four startup ideas to build on one, and covers getting customers beyond Meta's directory.

Zach Holman

@holman

An oldie post from 2024, but I've been curious about the data side. I had @Muse pull in my whole angel portfolio via MCP from @signedco of ~150 startups; the interesting bit: - Startups who sent investor regular updates: 8% died - Startups who didn't send updates: 28% died

Alex Kwon

@startupoppa

so i asked @muse and instinct book a restaurant and pay the reservation fee at the same time: - muse got through all far faster - but i went with instinct this time instinct let me use external browser browser so @link information was pre populated. so out of laziness i went