Featured · AIOps
Agentic AI in production: what actually works in 2026
The gap between an agent demo and an agent in production is enormous. After a year of shipping agentic features, these are the patterns that actually hold up.
EdWagon Blogs
Notes from the
Notes from the
AI-first classroom.
Field notes from practitioners — how teams ship AI products, what AIOps looks like in production, and the career playbooks that actually work in 2026.
Latest
Engineering
Small language models are eating the easy 80%
The interesting question in 2026 isn't 'which model is smartest?' — it's 'how little model can I get away with?' For most tasks, the answer is surprisingly little.
EngineeringReasoning models and test-time compute: the new scaling axis
For years, 'make it smarter' meant 'train a bigger model.' Now there's a second dial: let the model think longer at inference. It changes the economics of hard problems.
AI PMRAG isn't dead — it grew up into context engineering
Every few months someone declares RAG dead because context windows got bigger. They're missing the point. The work just moved up a level.
AI PMAI evals are the new unit tests
If you can't tell me how you'll know the model is doing its job, you can't ship it. Evals are how AI teams move fast without breaking things.
EngineeringThe inference cost curve: cost-per-token as a product decision
Every AI product decision is also a cost decision. The teams that treat inference spend as a design constraint from day one are the ones with a business at scale.
EngineeringOpen-weight vs closed models: how to actually choose in 2026
The honest answer is 'it depends' — but it depends on a small number of things you can actually reason about. Here's the framework we use.
EngineeringOn-device and edge AI: the quiet shift
The loudest AI story is the biggest models. The quietest — and maybe more consequential for products — is that good models now fit on the device in your pocket.
CareerThe AI engineer skill stack for 2026 (and how to build it)
If you're studying AI right now and wondering what's actually worth learning, this is the honest map — the skills hiring teams test for, in the order that compounds.
AI PMAI governance and the EU AI Act: what builders need to know now
You don't need to become a lawyer. But if you ship AI in 2026, 'we'll deal with compliance later' is no longer a strategy. Here's the builder's version.
