Each pioneer’s portrait is drawn by the technology of their phase — pencil, then characters of code, then pixels, then a relief of pure depth, then a mind wiring itself together from noise. The medium is the message.
The first system to cross the line sets the norms every later system inherits — openness, oversight, who it answers to. Defaults are destiny.
A frontier you can read, fork, and audit is one a whole community can catch failing. Secrecy doesn’t make intelligence safe; it makes it unaccountable.
No single lab should hold the steering wheel of the most consequential software ever written. So we’re building it in the open, with you.
infoRendered by hand, in pencil. The pre-digital craft, before there was a machine to draw with.
Before intelligence, infrastructure. CleverThis is the command center for the agent era — develop, version, deploy, and orchestrate any agent, local or cloud, in hierarchies built for the long, multi-step jobs that flatten single-shot models. This is the tape every later phase writes on.
Turing asked when we should grant that a machine thinks. We ask it on purpose, up front: every agent here is reproducible, inspectable, and open. You cannot align what you cannot read — so we make the whole stack readable while it is still safe to.
while (!halts) { instrument(); } // we can’t prove every run terminates — so we watch every run that doesn’t.infoRendered in ASCII — his face read as characters on a screen. Fitting, for the man who taught us that everything, even a face, is information.
You cannot improve what you cannot measure, and an agent’s behavior is just information. This phase turns runs into signal: evaluation harnesses as first-class artifacts, telemetry, long-term memory, and context compression. We score capability, faithfulness, and drift in bits — not vibes.
Hallucination is noise on the channel between intent and act. We drive it toward its entropy floor and defend the faithful signal — because a system you can’t trust to report itself accurately can’t be governed, no matter how capable.
SNR ≥ 0 dB or it does not ship. // Shannon juggled on a unicycle; we settle for juggling correctness and capability.
infoRendered in 8-bit — quantized to the few values an early memory could hold. The first machines that learned to see did it a coarse pixel at a time.
Agents stop being only what you configure and start adapting to what they do. Outcome-driven learning runs across the hierarchy; opt-in federated improvement learns from the whole community’s runs without hoarding anyone’s data. The platform begins to perceive.
This portrait is built from the corpus, tile by tile — and so is every model. Provenance, consent, and bias are not afterthoughts here: every contribution is opt-in, sourced, and revocable. We learned from the first AI winter that ignoring a system’s limits is how you freeze the whole field.
XOR is not linearly separable. Neither is ethics. // mind the second layer. (Minsky & Papert, 1969 — noted, not repeated.)
infoRendered as a ridgeline relief — every horizontal scan lifted by brightness, the face read as pure depth. The machine now perceives in three dimensions.
Recursive hierarchies run many levels deep; agents compose, supervise, and spawn other agents; general capability starts to emerge — behavior no one programmed line by line. This is the phase where power begins to outrun intuition, and where caution stops being optional.
Hinton walked away from a frontier lab to say this plainly: the risk is real. So we build the brakes before the engine — capability ceilings, mandatory interpretability, reversible everything, and an independent safety team that can stop a launch. The people racing fastest should be the ones most able to halt.
∂L/∂w propagates backward; responsibility propagates forward. // 2012: the year the cat could be recognized. 2030: the year we make sure it asks permission.
infoA self-organizing constellation — thousands of nodes scatter, then wire themselves into a face. Intelligence assembling itself from noise. (Watch it form.)
Recursively self-improving, aligned intelligence under genuine human stewardship: an explosion that is open, distributed, and value-locked to humanity. Not a product that is owned — a commons that is tended. Good’s caveat is the entire wager, and we intend to honor it.
That is Good’s caveat from 1965 — and it is the whole game. The first true AGI sets the defaults for every AGI after it: its values, its openness, its relationship to the people it serves. That is precisely why it must be open, auditable, and aligned — and why getting there first is not ambition, it is responsibility.
“The first ultraintelligent machine is the last invention that man need ever make.” — I. J. Good, 1965. // No basilisks were appeased in the making of this roadmap.
Not a vibe — a plan. Ship dates, safety gates, and the doublings between them. Every guard rail is a milestone, not a footnote.
Ambition without constraints is just risk with good PR. These bind every phase above. We publish them so you can hold us to them — and fork us if we don’t.
CleverThis launches this week — Phase I, in your hands. The other four phases get built in the open, by people who’d rather be in the room than read about it later. Bring a PR, a critique, or just your scrutiny.
psst — open the console. there’s a note for the curious. terminal
CleverThis is a sustainable AI gateway with hosted Actor endpoints — by the team behind CleverThis.
Join the newsletter
Product updates and engineering notes. No spam, ever.
