Publications from the Carpathian team
Research papers, release notes, security write-ups, and company news. Everything we publish about the infrastructure we run, the models we build, and what we learn operating both.
Featured
Free Unlimited AI Chat: No Caps, No Limits
Chat with a capable AI model for free on Carpathian. No message caps, no token limits, no metering on chat.
Read the article
Veritate, Carpathian's AI Inference Engine
Veritate is an AI inference engine that runs and trains large language models. It is tuned to work on a normal desktop CPU instead of a high-end graphics card.
AI Pen-Testing: What You Need to Know
Anthropic reports the first documented AI cyberattack. Chinese state-sponsored hackers use Claude AI to perform reconnaissance, exploitation, and data theft
All publications
Release Notes 2026.8.0
This update brings a fully redesigned website, Veritate AI, free threat intel feeds, Spark snapshots/restores, and billing.
Self-Hosted LLM vs API: A Cost Breakdown
Self-hosting an open-weight model or calling a hosted API? The drivers on each side, the break-even point, and the hidden costs.
What Building Infrastructure Taught Us About Trust
Years of building infrastructure taught me that trust is earned quietly, honesty about limits beats big promises, and locking people in is a slow death.
How to build semantic RAG AI search
A walkthrough of how to build semantic search with embeddings using RAG AI to index and chat with your docs, and when keyword search still wins.
How to use AI to pull data from images and PDFs
How to turn scanned images and PDFs into typed fields your app can store, using local OCR and a language model, with the validation that keeps poor quality out.
How to build an AI that ranks content appeal
We built a tool that scores titles and headlines with an LLM. Here is how to make the model grade consistently, hold a strict format, and stop it refusing.
Why software development teams slow you down
The textbook development lifecycle was built for a different era. Most of its ceremony delays the thing that matters.
Why We Self-Host Our AI: The Case for Owning Your Infrastructure
We chose to run our own AI infrastructure instead of building on closed APIs. Here is the case for owning your stack, and the honest cost of doing it.
Why Efficient Inference Beats Brute Force
I am convinced the future of useful AI is small, efficient, often task-specific models, not an endless race for the biggest one. Quality per watt beats brute force.
How to Add AI Content Moderation to Your App
AI content moderation done right runs cheap rules first, let a model judge the gray area, and flag the high-stakes calls for a human. Here is how we built it.
Using AI to help developers, not replace them
The useful question about AI in software is not how many engineers it replaces. It is how much faster it makes the good ones. Here is how we put it to work.
The Hidden Energy Cost of AI, and What Low-Power Inference Could Look Like
The AI energy cost is treated as an inconsequential while the power demand is at the size of some countries and why efficiency, not scale, is the future.
Why Data Quality Decides Small-Model Quality
On saturated consumer hardware, you do not buy small-model quality with a faster framework. You buy it with better data and a better training objective, with numbers to back it.
LLM Decoding Without Changing Output Bytes
Speculative decoding and multi-token prediction cut byte-level generation latency without touching output quality. Here is what we measured on Veritate's models
Why Low-Bit QAT Needs to Happen In Training
Below roughly four bits per weight, post-hoc QAT on byte-level models is catastrophic, and quantization-aware training recovers nearly all of the lost quality
Making an LLM Efficient By Optimizing Thinking Time, Not Parameter Count
For a model that must fit on a small device, the parameter budget is fixed. The place to spend is inference-time compute.
Upcycling the Data Center: Doing More AI With Older Hardware
We throw away a staggering amount of perfectly useful compute. Here is why a lot of AI and cloud work can run on older, upcycled hardware, and where it honestly cannot.
Managed vs Self-Hosted Databases: Which Should You Pick?
Managed or self-hosted? How to choose, compared on cost, maintenance, backups, and uptime, with the honest middle ground and a decision checklist.
IaaS vs PaaS vs SaaS, Explained Simply
IaaS, PaaS, and SaaS in plain terms: what the provider runs, what you run, examples of each, and how to choose the model that fits your project.
Data Sovereignty Matters More Than You Think
A region setting tells you where your data sits. It does not tell you whose laws can reach it. Why I think sovereignty is about to become a front-line concern.
Long-Context Memory For Constant-State Models
The conversation-needle benchmark, the transfer-gap finding, bitwise-exact streaming state carry, and the open programme for training past-window recall.
How A 121M Byte Model Learned To Talk
The three-phase recipe behind chat80m, the first conversing byte-level model in the project, its knowledge wall, and the identity-dosing discovery.
What Is a CDN, and Do You Need One?
What a CDN is, how edge caching speeds up your site and cuts server load, and the honest answer on when a small site needs one and when it does not.
The Composed Efficiency Stack
Byte-level training efficiency on one consumer machine: Muon, boundary patching, constant-state recurrence, and the hybrid trunk that beat both of its parents.