Lumabri
What if LLMs worked like Napster?
Details
- External ID
- 49236781
- Source
- HN
- Company
- —
- Product
- Lumabri
- Website domain
- github.com
- Launched
- Aug. 9, 2026
- Cohort
- —
- Upvotes
- 9
- Upvotes percentile
- 0.5309139784946236
- Tags
- —
- Fetched at
- Sept. 10, 2026, 5:32 a.m.
- Updated at
- Sept. 10, 2026, 5:32 a.m.
Description
A while ago I started working on Colibrì to see if it was possible to run huge LLMs on a normal computer. The project grew far beyond what I expected, thanks in large part to the HackerNews community.That led me to a new question:What if we stopped thinking about one computer?This is the idea behind Lumabri.Instead of requiring a single machine to store and run an entire huge model, Lumabri treats a network of normal computers as a shared pool of resources.One machine might provide disk space, another compute, another a different part of the model. If a required block or expert isn’t available locally, the system can retrieve or execute it on a peer.This is particularly interesting for Mixture-of-Experts models. A model can have hundreds of billions of parameters, while only a fraction are activated for each token. Rather than moving huge expert weights over the network, Lumabri can send the small activation to a peer that already has the expert and let it execute it.The goal is for machines to contribute whatever resources they can afford while using the swarm for the rest.The idea is very much inspired by peer-to-peer systems: users are the infrastructure.There are obviously major challenges, especially network latency and security. I’m experimenting with peer verification, SHA-256 verification, signed model state, replica selection, failover, and deterministic execution.Lumabri is still an early experiment. I don’t have a datacenter or a huge GPU cluster, so I’m building it with the hardware I have and trying to find out whether the idea actually makes sense.With Colibrì I asked:Can one normal computer run a huge LLM?With Lumabri I’m asking:What if many normal computers could become one huge computer?Feedback welcome.Repo: https://github.com/JustVugg/lumabri
Enrichment
- Theme
- lightweight and on-device AI runtimes
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- AI-native
- Project type
- Commercial product
- Normalized one-liner
- distributed llm inference network
- Manually corrected
- False
Could you build this?
No Lumabri (and Colibri) is a peer-to-peer distributed LLM inference network running fragmented model layers across consumer computers over the internet (reminiscent of Petals/Napster). Building distributed peer-to-peer tensor parallelism over high-latency, heterogeneous consumer connections requires advanced distributed systems and deep learning systems research.
What it would actually take: The architecture requires custom P2P networking (e.g. libp2p), pipeline/tensor model parallelism across untrusted consumer nodes, custom dynamic routing to handle churn and node latency, and quantization kernels (CUDA/Metal). The main bottleneck is coordinating multi-gigabyte activation transfers over variable WAN bandwidth while ensuring Byzantine fault tolerance and incentive alignment. This requires specialized distributed systems researchers and ML infrastructure engineers.
Discussion
10 comments analyzed.
Competitors mentioned: Bittensor, Federated learning systems, Cryptocurrency verification approaches
Concerns raised: Network latency makes token generation hundreds of milliseconds per token, borderline unusable, Model weights too large to fit on consumer GPUs even high-end cards, Security verification difficult in P2P - peers can lie about calculations, Privacy risks from logging personal queries to peers, Potential for misuse if AI output used for harmful purposes
Feature requests: Bond/forfeit system for peers to prevent lying about calculations, Security safeguards against malicious query execution, Clearer readme documentation without excessive information
Competitors
Other products that read as similar to this one — 59 launches clear the similarity bar, closest 8 shown.
Attention rank: #31 of 60 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 279 days after the earliest competitor.
- Getting GLM 5.2 running on my slow computer · hn · 2026-07-09 · 937 upvotes · similarity 0.55
- Lume 0.2 · hn · 2026-01-18 · 154 upvotes · similarity 0.43
- Llm.sql · hn · 2026-04-24 · 8 upvotes · similarity 0.41
- Tokenflood · hn · 2025-11-12 · 21 upvotes · similarity 0.41
- Run 500B+ Parameter LLMs Locally on a Mac Mini · hn · 2026-03-09 · 17 upvotes · similarity 0.40
- LUML · hn · 2026-02-03 · 7 upvotes · similarity 0.38
- We built a <60ms, open-source alternative to E2B using RustVMM and KVM · hn · 2026-04-22 · 7 upvotes · similarity 0.37
- Clone, a small Rust VMM, forks VMs in under 20ms via CoW · hn · 2026-04-19 · 11 upvotes · similarity 0.37
Other launches for this product
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.