KI Norge trenger fart, kontroll og ryggrad
Lanseringen av KI Norge ga meg både mer håp og mer uro. Håp fordi regjeringen endelig snakker om fart, samordning og ansvarlig bruk i samme åndedrag. Uro...
Tip #44: Steer your agent from your phone
Both Copilot CLI and Claude Code let you steer a running agent from your phone — approve edits, send prompts, check results while AFK.
Three Hooks That Give Claude Code Memory
Three passive hooks that give Claude Code memory, routing, and hygiene — without changing how you work. The session amnesia problem, and how to fix it at the...
10 spørsmål alle ledergrupper bør stille om AI før sommerferien
Kunstig intelligens handler ikke lenger om å følge med på en teknologitrend. Det handler om hvordan virksomheter skal lede, organisere, lære, konkurrere,...
Fremtidens konkurransekraft blir digital, grønn og sikker samtidig
Norske virksomheter kan ikke lenger behandle digitalisering, bærekraft og sikkerhet som tre separate prosjekter. Kundene, myndighetene, leverandørene,...
Production Scars Are Architecture
Six architectural lessons from running an autonomous AI agent in production. Each scar is a failure category surprising enough to force mechanical change....
Your AI Agent Does Not Know the Law (and How to Fix That)
Your AI agent will confidently tell a customer they're GDPR-compliant when they're not. Here's the architecture that fixes that.
Tip #43: How to write your first skill
Two paths to writing skills: translate existing runbooks or ask the agent to draft from a successful trace. The description field is the routing algorithm....
The Zombie in the Basement
Static architecture catalogs document what someone knew at a point in time. Betonix GmbH thought they had 23 database connections. They had 52. The migration...
Ansvarlig AI vil bli et konkurransefortrinn
Ansvarlig bruk av kunstig intelligens blir ofte fremstilt som noe som bremser utviklingen. Jeg tror det motsatte kan bli tilfelle. Virksomheter som bygger...
Google's Open Knowledge Format and the problems it deliberately doesn't solve
Google's Open Knowledge Format standardises content packaging for agents but stops before trust, temporality, and governance. KCP covers exactly that territory.
Hvordan påvirker AI hvordan vi jobber med dataanalyse?
I denne episoden deler to av våre dataanalytikere hvordan kunstig intelligens i løsninger som Microsoft Fabric, Power BI er i ferd med å snu opp ned hvordan...
De beste arbeidsgiverne bruker AI til å styrke de ansatte
Kunstig intelligens kan bli et kraftig verktøy for effektivisering, men virksomheter som bare bruker AI til å presse mer ut av færre mennesker, risikerer å...
Tip #42: The advisor pattern — frontier quality at fraction of cost
A small, fast model executes; a powerful model sits as a callable advisor. Eve Legal: £110 → £11 per run (10×). The structure of your harness matters more...
Sixteen Versions of Metadata Nobody Read
63 regulatory manifests with full KCP v0.21 metadata -- hashes, signatures, temporal validity. Then a benchmark showed agents were not reading any of it.
Discoverable Is Not Navigable
Discoverable is not navigable. Extracting 38 Dutch obligation units from inline YAML to standalone text made content agent-traversable, not just findable.
Stale Knowledge Is Worse Than No Knowledge: KCP v0.19 and v0.20 Close the Temporal Gap
Stale knowledge is worse than none. KCP v0.19 adds bi-temporal validity and supersession links; v0.20 adds as_of queries for point-in-time compliance.
2026 06 12 kcp 0.18 unit integrity
KCP v0.18 adds per-unit content hashes and origin evidence. Signing the manifest proves the map is authentic; hashes prove the territory matches.
Down the Rabbit Hole: How a 33-Tool-Call Bug Became a Knowledge Standard
KCP started when a 33-tool-call bug revealed agents waste their budget on discovery. Six months later: a knowledge standard with trust and governance.
Beyond RAG: How KCP 0.16--0.17 Give Agents Trustworthy, Self-Describing Knowledge
KCP v0.16 adds cryptographic signing and trust tiers. v0.17 adds content structure metadata and subtractive audience filtering for precision retrieval.
Konsulentbransjens juniorproblem
AI gjør konsulentene raskere. Det gjør dem ikke nødvendigvis mer verdifulle. For dagens unge konsulenter kan kunstig intelligens bli en historisk...
Explorative Development
Before writing production code, run a fictional organisation through two years of using the feature. Front-load discovery before the first commit.
Organized Truths
Catch an agent making a false compliance claim once, encode the verification as a skill, and every future session inherits the check. Organised truths scale.
False Alarms and False Assurances
An agent claimed an API had no authentication. Four minutes of reading proved otherwise. Verifying agent assertions is the non-negotiable skill.
Et stort stykke arbeid for et lite stykke Norge
Helgeland er en vakker region med ca. 85.000 innbyggere, på størrelse med Oslo-bydelene Sagene og St. Hanshaugen til sammen. Den minste kommunen har kun 463...
KI-skandalene avslører et større problem i offentlig sektor
Tromsø og Asker kommune har vist oss hvor galt det kan gå når kunstig intelligens brukes uten god nok kontroll i offentlig saksbehandling. Men den mest...
Tip #41: Agent costs rise - should you care?
AI tool pricing has changed — Copilot is now usage-based, Claude Code has premium tiers. Most users are fine, but heavy agent users should check their spend...