LinkedIn·Monday, 17 August 2026·9d ago
Hot take: a stronger AI model doesn’t automatically produce better analytics. We have seen stronger models: > take longer to answer > use…
Karel Callens
CEO & Founder at Luzmo
Hot take: a stronger AI model doesn’t automatically produce better analytics.
We have seen stronger models:
> take longer to answer
> use more tokens
> question decisions that didn’t need to be questioned
> produce less reliable results for a specific task
The best model on a benchmark may not be the best model inside your product.
For many workflows, a more deliberate setup works better: use an expensive model where planning or difficult reasoning matters, then use faster, cheaper models for well-defined execution.
More intelligence is useful.
More intelligence applied indiscriminately can become overhead.
♥ 8
View on LinkedIn Cross-referenced
Related on the wire
I DON'T WANT every useful analytical answer to look like a Luzmo dashboard. Strange thing to say as Luzmo's CEO. But sometimes the right…
We use BrightAnalytics to follow our financial numbers. It's the authoritative source, and it's good at what it does. But I don't want to…
We're hiring a Business Developer for our telecom vertical. Telecom operators sit on anonymized network location signals that are valuable…
You: Can we build this ourselves? AI: Done. First dashboard is ready. A month later: You: Can we connect it to live data? AI: Connected.…
In 2019, I expected voice interfaces to change analytics MUCH faster than they did. There were plenty of voice startups and the tooling…
We're growing in the US and hiring in New York. Noah Fauchon is building something special out of our DUMBO office. If you know a great AE…