The Cheapest Model Is Not the Cheapest Model

We built an evaluation harness around a production agent, tested models on the workload it performs, and cut inference cost by 86% while preserving…

Sep 10 • Alf Viktor Williamsen

1

1

When Does Owning Inference Pay?

An API is a good place to start. Owning the hardware begins to make sense when the workload, or the need for control, justifies the bill.

Sep 9 • Alf Viktor Williamsen

1

2

The Token Is Not the Unit of Intelligence

The economics behind the bill, and why tokens are becoming an incomplete measure of intelligence.

Sep 7 • Alf Viktor Williamsen

4

3

Crown Chasers and Price Breakers

The game theory of the AI labs.

Sep 6 • Alf Viktor Williamsen

1

1

August 2026

Introducing Entopic: Deep Research for Hiring

Build the best possible talent pool. Make calls with research-grade confidence.

Aug 30 • Alf Viktor Williamsen

1

March 2026

Your app should have an email address, not just a login page

Think about how a traditional SaaS app works.

Mar 15 • Alf Viktor Williamsen

2

1

Giving an LLM Full Database Access Is the Right Call, and Here’s the Specific Design That Makes It Safe

One of the most exciting applications of LLMs I’ve worked with is giving an agent full access to run commands on a database.

Mar 6 • Alf Viktor Williamsen

1

Data That Answers When You Ask

Businesses collect data perpetually: product types, customer profiles, refund rates, currencies, sales identifiers, personnel records.

Mar 3 • Alf Viktor Williamsen

1

1

February 2026

Production PDF Parsing for High-Stakes Financial Systems

Reading PDFs is a vital function of any application dealing with documents.

Feb 22 • Alf Viktor Williamsen

2

1