// the manifesto
Why we watch the road.
Toward AGI is an independent publication covering the systems, labs, and ideas on the road to artificial general intelligence.
Somewhere between the press releases and the doomsday threads, there is an actual story: a technology advancing faster than any before it, built by a few thousand people, watched by everyone, and understood — fully — by almost no one.
We started Toward AGI because the coverage of that story was broken. Too much of it is marketing republished as news, or fear repackaged as analysis. We wanted a third option: dispatches written by people who read the papers, run the models, and still have the patience to explain things properly.
The name is a direction, not a prediction. Whether AGI arrives in five years or fifty, the road there is the defining technical story of this era. Every model release, every eval, every policy fight is a mile marker. We log them all.
// operating principles
01
Attention is the product
The most important technology of our lifetime is being built in public. Our job is to watch closely — and to tell you what we saw without the hype distortion field.
02
Plain language, no dumbing down
We explain attention, RLHF, evals, and scaling so a smart non-specialist can follow — without pretending the hard parts are easy.
03
Receipts or it didn't happen
Claims get linked. Benchmarks get sourced. When we speculate, we say so in plain text.
04
The radar never sleeps
An automated sweep of Hugging Face and OpenRouter runs every six hours. If a model ships, it lands on the radar — usually the same day.
// what we cover
News
Fast, factual reporting from the frontier — funding rounds, lab moves, policy shifts, and everything that moved the needle this week.
Model Releases
First looks at new models, captured by our radar within hours of release. What shipped, what it claims, and what actually matters.
Trends
Pattern recognition at scale — where the field is heading, who is converging on what, and which bets are compounding.
Education
Concepts explained properly. From attention to RLHF to eval design — written to be understood, not to impress.
Deep Dives
Long-form technical investigations. We take one idea apart, bolt by bolt, and put it back together in front of you.
Applied AI
AI in production — deployments by corporations, institutions, and governments, plus novel applications solving real-world problems.
“Someone should be paying attention — carefully, critically, and in plain language. So we are.”
Join the publication →