LLM stories
Large language models have moved from research curiosity to production infrastructure. This topic focuses on LLM APIs, fine-tuning, RAG pipelines, agent frameworks, and the economics of running models at scale.
HN discussions here tend to be unusually practical — benchmark comparisons, cost breakdowns, and real deployment war stories from teams shipping LLM features.
895 stories archived · Page 14 of 30
RSS feed for LLMGiving an LLM your prod database is easy. Taking access away is the hard part
deepsql.ai · 4 points · 6 comments
Direct light-to-token conversion with integrated 2D photosensitive memory
nature.com · 4 points · 0 comments
Why Elasticsearch is becoming a columnar database
elastic.co · 17 points · 1 comments
Run 290B+ frontier MoE models locally on your gaming PC
github.com · 36 points · 3 comments
Declarative-forms – await an object the way prompt() awaits a string
wolfoo2931.github.io · 13 points · 1 comments
LFM2.5-DSpark: Up to 3.2x Faster Inference from H100 to MacB
liquid.ai · 18 points · 0 comments
LiteLLM (YC W23) Is Hiring – Rust / Performance Engineers
jobs.ashbyhq.com · 1 points · 0 comments
Steganeur – Hide secret messages in LLM-generated text (Rust)
github.com · 3 points · 0 comments
LLMs are (still) mostly powered by imitative learning, not RL
lesswrong.com · 3 points · 0 comments
Bounded GenAI metrics from Otel traces with O raw prompts
github.com · 4 points · 2 comments
Good Results when training Qwen 3 4B to learn a new domain
teachmecoolstuff.com · 5 points · 0 comments
The disjointed Prague of Deus Ex: Mankind Divided
killscreen.com · 29 points · 10 comments
We burned 11.7B tokens to find the best cyber AI model
aikido.dev · 14 points · 6 comments
Let your coding agents set up your test coverage
blog.codacy.com · 2 points · 0 comments
Single-User Inference Card
markohaberl.substack.com · 3 points · 0 comments
Teaching a local LLM to reason about a new domain through continued pretraining
teachmecoolstuff.com · 6 points · 0 comments
How do you review and validate LLM generated code?
4 points · 2 comments
America's Colleges Are Hurtling Toward an Enrollment Cliff
wsj.com · 15 points · 18 comments
Google Cloud us-west1 down
11 points · 4 comments
Clean up Claude 5's token vomit with a separate LLM
github.com · 65 points · 299 comments
SanDisk Tapes Out Its First HBF Memory Die, Targets 2027 for Product Samples
storagereview.com · 3 points · 0 comments
LLM-enabled large data analysis
app.verbagpt.com · 2 points · 0 comments
Ullis – Local Ternary Moe-Kan Training and Inference Engine in Rust
github.com · 3 points · 0 comments
Guess which of these LLM outputs is watermarked
sgoedecke.github.io · 11 points · 85 comments
Every Model Cheats
dreadnode.io · 23 points · 97 comments
Parmar – BPE tokenization as a pre-filter for LZMA (+9.6%, and faster)
github.com · 4 points · 0 comments
Find every AI model your code calls and warn before it's retired
llmstatus.ai · 5 points · 1 comments
Do Chatbot LLMs Talk Too Much?
arxiv.org · 12 points · 4 comments
Idea to reduce AI token use at large orgs
3 points · 0 comments
Sentrint, a security scanner for projects build with LLMs
sentrint.com · 6 points · 2 comments