Advanced tag | eYssen AI

25 Articles

Advanced ×

Close to twenty AI tools running side by side, and whatever one of them learned, the next had never heard of. That is where EYAS came from. This post is not a feature list: it shows the decisions, the trade-offs we said out loud, and the one measurement that threw a finished-looking solution out of the window.

Advanced Business owner Case study EYAS Agents & Automation

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Fine-Tuning: When Is It Finally Worth It?

A fine-tuned model with 350 million parameters hit a **77.6%** success rate on a tool-calling task — while a general model with 175 billion parameters managed **26%** on the same task. Five hundred times fewer parameters, three times the result. That is the real promise of fine-tuning. And meanwhile one large provider is **retiring its fine-tuning platform**: no new users are being accepted, and existing ones have until January 2027.

Advanced Developer Explainer Data, RAG & Self-Hosted AI

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Private AI: Keeping Data Inside Your Company

One large provider's zero data retention agreement lists item by item which endpoints it covers — and which it does not. On the does-not list: **file upload** and the **vector store**: **exactly the two features you would use to build RAG on that platform.** And on the other side, the assumption that "local means safe": a 293-day survey found **175,108 publicly reachable local model servers** in 130 countries.

Advanced Business owner How-To Data, RAG & Self-Hosted AI

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Quantization: How Big Models Fit on Small Machines

According to a unified 2026 measurement, four-bit quantization **cuts model size by 71%** and costs on average **0.43 percentage points** of accuracy. Essentially free. Except **not for everyone**: the same four-bit quantization also produced a **59%** collapse on long input, and the degradation is **up to five times worse in languages other than English** — measured across 26 languages, Hungarian among them.

Advanced Developer Explainer Data, RAG & Self-Hosted AI

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Hardware for Local AI: VRAM, RAM and Apple Silicon

The rule of thumb everyone repeats says you need about half the parameter count in memory. `[Own calculation]` The four-bit weights of an eight-billion-parameter model really do eat 4.6 gigabytes — **but the same model's 128k-token context eats 18 gigabytes**, four times the model itself. Anyone who only looks at model size has worked out a fifth of the memory they actually need. And there is one measurement that turns the whole picture around: **a chip with half the GPU cores processes a prompt 83% faster than a more expensive one** — while generating a fifth slower.

Advanced Developer How-To Data, RAG & Self-Hosted AI

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

How Do You Know Your RAG System Is Right?

At a public evaluation campaign a judge model was measured against a human assessor. Their agreement, corrected for chance, came out at **0.29**. On the very same task, **two humans agreed at −0.03** — worse than chance. So this is not a story of "the machine is bad, the human is good". The task itself is ill-defined at the edge cases. This article is about what can still be measured reliably.

Advanced Business owner How-To Data, RAG & Self-Hosted AI

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Hybrid Search: Why Vectors Alone Are Not Enough

In an academic measurement of 24 thousand questions, vector search scored **19.2%** on questions of the "who is [X]'s child?" type — the decades-old keyword algorithm scored **82.9%**. Sixty-three points apart. On a different set of questions the very same methods come out the other way round. **Neither one is better than the other: they have different failure modes.** Except that hybrid search is not what improves things most — re-ranking is.

Advanced Developer How-To Data, RAG & Self-Hosted AI

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Preparing and Chunking Your Documents

In one measurement, simply cutting a long document into chunks lifts retrieval from **28.2 to 45.7** — **17.5 points for free**. The same measurement also shows that **bigger chunks are monotonically worse**. The awkward part: the most widely used tools ship with defaults of 800 and 1,024 tokens, or 4,000 characters — and in an independent test the 800-token setting finished **near the bottom of the field**.

Advanced Business owner How-To Data, RAG & Self-Hosted AI

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Project Instructions an Agent Actually Follows

In a paired experiment, a project instruction file meant **20% fewer tokens and 20% less time** for the same task. What nobody has demonstrated is that it makes the code better. And there is a ceiling: **at roughly eighty simultaneous rules**, "follows every rule" essentially stops happening.

Advanced Developer How-To Coding & Development

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

The Security Bill of AI-Generated Code

From 11 September 2026, anyone placing software on the market in the European Union **has to report an actively exploited vulnerability within 24 hours**. Meanwhile roughly **45%** of generation tasks produce code with a known vulnerability when there is no explicit security instruction. Those two numbers together are the bill.

Advanced Developer Explainer Coding & Development

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

Hallucinated Packages: When AI Invents a Library

**43%** of invented package names come back identically in all ten runs out of ten. That is what turns a hallucination into an attack surface: the attacker does not have to guess what the model will invent — one question is enough. A documented, successful attack, on the other hand, **had not happened** by mid-2026 — and that honest answer is stronger than scaremongering.

Advanced Developer Troubleshooting Coding & Development

Sep 13, 2026

Krisztian Eyssen, Krisztian Eyssen

AI in Code Review: Which Bots Are Worth It

In an independent measurement, the precision of AI review agents is **3.2–5.1%** — roughly one comment in twenty lands. In a peer-reviewed vendor study, meanwhile, **38.7%** of the comments triggered a code change and PR turnaround got **30.8%** shorter. Both are true. They are not measuring the same thing — and that is what this post is about.

Advanced Developer Tool review Coding & Development

Sep 13, 2026

Tags

Audience

Everyone Office worker Marketer Creative Business owner Developer

Follow Us

Archives

Tags