Archive / All postsNewest first

All posts

Every post and tracker on AI agents, GraphRAG and retrieval, AI transformation and what AI costs, newest first.

TrackerUpdated Oct 29, 2026
EU AI Act dates and guidance, tracked
Which EU AI Act obligations apply when, after the 2026 AI Omnibus moved the high-risk deadlines, plus harmonised standards and fines. Updated when a date or a standard changes.
Ivan Bardak
- Governance
- Regulation
TrackerUpdated Oct 29, 2026
AI agent reliability benchmarks, tracked
How reliably the best AI agents succeed when it counts, in plain language: success on every try, the length of task they can handle, and realistic office work. From the official leaderboards, updated when they change.
Ivan Bardak
- Benchmarks
- Evaluation
TrackerUpdated Oct 29, 2026
MCP specification changes, explained
What each version of the Model Context Protocol changed, in plain language: how users sign in, what tools may do, and what to check in your own MCP servers. Updated when a new version is published.
Ivan Bardak
- MCP
- Security
Oct 29, 202613 min read
Agentic AI from demo to production: what the evidence says about the architecture
Most agent pilots stall on the architecture around the model, not the model. What surveys, published incidents and reliability benchmarks show about tools, identity, retrieval, reversibility and measurement, and the six questions I ask before anything ships.
Ivan Bardak
Oct 29, 20266 min read
GraphRAG vs RAG: what the evidence actually says
GraphRAG wins on multi-hop and whole-corpus questions and loses on plain lookups, at a higher cost. What three studies show, how much weight each deserves, and a decision rule you can use.
Ivan Bardak
Oct 29, 20266 min read
Five questions before you build a knowledge graph for RAG
A knowledge graph can make RAG better at connecting facts, or just slower and more expensive. Five questions, each tied to published evidence, that tell you which one you are about to get.
Ivan Bardak
Oct 29, 20266 min read
A GraphRAG business case, done the evidence-based way
Most GraphRAG business cases are a vendor benchmark and a hopeful percentage. A structure that survives a CFO's questions: six steps of evidence-based management, a cost model from measured token counts, and a value range instead of one number.
Ivan Bardak
Oct 29, 20267 min read
AI code migrations: what Airbnb, Google and Amazon actually reported
The headline numbers travel well: 1.5 years in 6 weeks, 50% less time, 4,500 developer-years saved. What each company actually measured, how much weight it deserves, and a decision rule for your next migration.
Ivan Bardak
Oct 29, 20266 min read
Does AI make developers faster? What the controlled studies say
55% faster, 21% faster, 19% slower, roughly even: the controlled studies disagree for good reasons. What each one measured, how much weight it deserves, and what to put in an AI transformation plan instead of a single number.
Ivan Bardak
Oct 29, 20266 min read
MCP tool design: fewer, safer, testable tools
Accuracy falls as the tool list grows, and every major vendor now says keep it small. What the MCP specification, the vendors and independent benchmarks say about tool design, and the checklist I use to review an MCP server.
Ivan Bardak
Oct 29, 20266 min read
Delegated access for agents: OAuth 2.1, PKCE and scoped tokens
On whose behalf is the agent acting, and with what rights? The delegation model I use, what the MCP authorization specification requires as of 2026, how to design scopes, and where agent credentials actually leak.
Ivan Bardak
Oct 29, 20266 min read
The real cost of GraphRAG: indexing, queries and keeping the graph fresh
Indexing reads your whole corpus with an LLM, global search can use hundreds of times more tokens per question than plain RAG, and freshness is the cost nobody budgets. Published measurements, current prices and a cost model you can fill in.
Ivan Bardak
- Enterprise
- GraphRAG
Oct 29, 20266 min read
AI transformation frameworks compared: which one is for what
The EU AI Act, ISO/IEC 42001, the NIST AI RMF and maturity models answer different questions. What each is for, what is mandatory, the AI Act dates after the 2026 Omnibus, and how I combine them in a programme.
Ivan Bardak
- Enterprise
- Governance
Oct 29, 20265 min read
"95% of AI pilots fail": what the MIT report actually measured
The most quoted number in AI strategy decks comes from a preliminary report built on 52 interviews, and it means four different things. What the MIT NANDA report measured, what it gets right, and what larger surveys say instead.
Ivan Bardak
- Enterprise
- Evaluation
Oct 29, 20265 min read
Klarna's AI customer service: the 700-agent claim and the U-turn
In 2024 Klarna's AI did the work of 700 agents; in 2025 the headlines said it went back to humans. What the SEC filings show: AI now handles 80% of chats, cost was the wrong main measure, and the model became dual-track.
Ivan Bardak
- Case study
- Enterprise

EU AI Act dates and guidance, tracked

AI agent reliability benchmarks, tracked

MCP specification changes, explained

Agentic AI from demo to production: what the evidence says about the architecture

GraphRAG vs RAG: what the evidence actually says

Five questions before you build a knowledge graph for RAG

A GraphRAG business case, done the evidence-based way

AI code migrations: what Airbnb, Google and Amazon actually reported

Does AI make developers faster? What the controlled studies say

MCP tool design: fewer, safer, testable tools

Delegated access for agents: OAuth 2.1, PKCE and scoped tokens

The real cost of GraphRAG: indexing, queries and keeping the graph fresh

AI transformation frameworks compared: which one is for what

"95% of AI pilots fail": what the MIT report actually measured

Klarna's AI customer service: the 700-agent claim and the U-turn