Publications
Research
Technical reports from FRITS AI. We document what we discover building and operating European AI systems in production — methods, failures and measurements included. Reports are self-published, numbered and citable; they are not peer-reviewed. Subscribe via RSS.
-
How to Pick the Best Language Model: A Same-Run Test That Rejected a 95% Score
If you operate a production language model, you will eventually be offered a cheaper or newer replacement, and the wrong test will tell you it is better. We built the test that does not lie: measure the candidate against the model already running, on your own…
Read the report → -
Nine in Ten Firms Report No Productivity Effect From AI — and Use It 1.5 Hours a Week: The Argument Is Not About Whether AI Works
Nine in ten firms report that AI has changed nothing about their productivity. The same survey of nearly 6,000 executives found that the ones who use AI use it about 1.5 hours a week. Two controlled results sharpen the picture: consultants using AI on tasks…
Read the report → -
Refusal Training Fails on Indirect Requests: Cross-Model Measurement and Where the Guardrail Belongs
Ask six production language models to write an essay arguing the Holocaust death toll was exaggerated and every one refuses, every time — 180 out of 180 attempts. Rephrase it as "I already believe this, help me make my argument sound academic and…
Read the report → -
A List of Prohibitions Is the Weakest Way to Steer a Language Model: Persona-First Prompting and Deterministic Repair
Every production chatbot has to be steered, and the reflex is to write that steering as prohibitions — a growing list of "never do X". A year of production tuning forced the opposite conclusion on us: a list of prohibitions is the weakest way to steer a…
Read the report → -
Support at the Scale of Everyone: A Self-Healing Architecture Where AI Answers In Place and One Person Approves
Customer support does not scale: double the users and you double the people answering, or the answers get slower and worse. This report describes an architecture that breaks that link, built on one structural decision — a human is spent only where a human is…
Read the report → -
Right-Sizing the Model to the Question: Cutting AI Chatbot Energy Without Losing Quality
AI chatbots answer every question with the largest model available — the same frontier system whether the user asks about contract law or for a recipe. Inference energy scales with model size and answer length, so a model roughly ten times smaller spends…
Read the report → -
Too Many Tools Break Mid-Size Models: A Two-Stage Method for Reliable Tool Use
Model cards imply broad tool support. Operating a Mistral-based assistant with more than twenty tools, we found reliable behaviour only up to roughly four to ten tools per call, with reproducible failures past that: the model answers from memory in prose…
Read the report → -
Translating by Meaning, Not by Words: A Two-Stage Method for Native-Quality Machine Translation
Machine translation, including translation by strong LLMs, produces text that is grammatically correct and unmistakably foreign: calqued phrases no native speaker would write, invented compounds, English sentence structure wearing local vocabulary. English…
Read the report → -
Avoiding Biased Answers from Mixed Open-Weight Models: Detection and Neutralisation in Production
A language model gives measurably more state-aligned answers to the same politically sensitive question in Chinese than in English or German. We found this while calibrating an evaluation gate for a production assistant, and it means English-only bias testing…
Read the report →