Swiftlet
AI
local inference
SHOW HN2026
How HN took it
8 positive, 18 neutral or asking, 0 mixed, 7 critical, out of 33 comments about the project. Thread score +0.01.
Performance: 4 for, 4 againstIdea: 2 for, 2 againstPricing: 0 for, 1 againstDocs: 1 for, 0 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Progress on on-device inference is exciting
From 2 comments
I know everyone wants to crap all over these setups that are impractical, but this is how progress happens. People will keep plugging away at this and figure out how to avoid…
dghlsakjg on HNRunning a 35B on iPhone at 1 tok/s with 2.5 GB RAM… this is the future of on-device inference. Insane work.
nc55g3g on HN
Useful project that makes big models run locally
From 1 comment
This looks useful, you can increase the RAM cache so if you have a Mac with 24-32GB it should speed up a lot and still run models that wouldn’t normally fit. I’m going to run…
adrianco on HN
Objections
Prefill is too slow for long prompts
From 2 comments
this is cool but like, are we just vibe coding NAND burners at this point? these decode times don't really tell the whole story, because prefill becomes the bottleneck. half an…
brrrrrm on HNagree, prefill is the weak spot right now. it goes through the same per-token path as decode, which is dumb for long prompts. The fix is on the list: during prefill we can batch…
leonickson on HN
SSD offload is slow and may wear out drives
From 1 comment
At what, 10 tokens per hour? These disk swapping methods all have the same drawbacks - kill your drive early, and slow as hell.
jbird99 on HN
Not practical beyond chat-length prompts
From 1 comment
> but this is how progress happens. This is progress in the same way that a man climbing a tree is making progress toward reaching the moon. This project is essentially the…
pizza234 on HN
Requests
Tunable RAM cache usage on larger-memory Macs
From 1 comment
I know relatively little about the workings of LLMs, but I keep seeing projects like this that run massive MoE models using very modest amounts of RAM, perhaps excessively so. I…
CyLith on HN
Support for Android, Linux and Windows
From 1 comment
Is there something that already runs like this on android / linux ( / windows)? (ollama or something?) Or could this be ported to work on other such platforms? edit: AI mentions…
erelong on HN
Support for other models such as Gemma
From 1 comment
How can I do this with, say, Gemma?
lenerdenator on HN
Questions
How does web access work with local models?
From 1 comment
The web and connecting to other services is very important for almost all of my use cases. While I believe we are going to get better and faster models, the web index is…
gizmodo59 on HN
Why was Claude Code credited in the README?
From 1 comment
I see this at the end of the README > Swiftlet was built in collaboration with Claude Code. Did this really happen (some sort of working with Anthropic or Claude Code team) or is…
crossroadsguy on HN
Alternatives named
BigMoeOnEdge
From 1 comment
See also: https://github.com/Helldez/BigMoeOnEdge
lern_too_spel on HN
TurboFieldfare
From 1 comment
Thank you for using TurboFieldfare as a starting point for this project and thank you for mentioning it at the README. I am glad it inspired more people to explore area of…
gitpusher42 on HN
Ollama
From 1 comment
Is there something that already runs like this on android / linux ( / windows)? (ollama or something?) Or could this be ported to work on other such platforms? edit: AI mentions…
erelong on HN