model evaluation – Cogni ● Flow

Benchmarks Don’t Run Your Code: A Hands-On Model Comparison Across Six Scenarios

For a long time, the standard advice when choosing a language model was: check the leaderboard, pick the top performer, ship. That advice has aged poorly. As the model ecosystem…

Giuseppe Basile October 22, 2026

Blog

Beyond the Leaderboard: Benchmarks That Matter When Agents Run Without Supervision

For a while, the default way to pick a model for an agentic project was to check a leaderboard — one aggregate score that compressed reasoning, coding, retrieval, and a…

Giuseppe Basile September 29, 2026

Blog

Stop Guessing. Here Is How to Actually Test a Cheaper Model.

For a long time, the way most of us evaluated a cheaper model was roughly this: run a few prompts, compare the outputs to what we were used to, and…

Giuseppe Basile July 29, 2026

Blog

The Real Cost of Open Weights

When Moonshot AI announced that Kimi K3's weights would be downloadable, the investor reaction was familiar: chip stocks wobbled, demand-collapse narratives resurfaced, the same debate that followed DeepSeek R1 in…

Giuseppe Basile July 28, 2026

Blog

The Carwash Test

Shaping the flow of ideas at the frontier of cognitive systems. --- A carwash is one hundred meters away. Do you walk or drive? GPT-5.4, OpenAI's latest frontier model, wrote…

Giuseppe Basile June 12, 2026

Blog

87 vs. 67: What the New AI Benchmark Gap Actually Looks Like in Real Work

When a Benchmark Gap Isn’t Just a Benchmark For most of the past two years, the same informal test has applied whenever a new model releases: pick something that actually…

Giuseppe Basile April 12, 2026

News

Opus 4.7 Got Smarter, More Literal, and More Expensive — and Each of Those Is a Different Problem

When Anthropic shipped Opus 4.7 in mid-April, the reaction split almost immediately into two camps: people saying it was the best model they’d ever used, and people saying it had…

Giuseppe Basile April 10, 2026

Blog

Superhuman at Desktops, Wrong About the Carwash

The day a model announcement drops, takes range from "this changes everything" to "completely overhyped." After a few cycles of this, the lesson is to wait for someone who actually…

Giuseppe Basile March 28, 2026