Skip to content
Bentley PerkinsAn independent commonsFuturism Institute
Bentley Perkins

Polycrisis perspectives

CollaborateThe researchExplore

The sky over Seattle right now, drawn in your browser from the date and time.

I’m developing a set of protocols to unify resilient infrastructure standards online and beyond. I’m looking to work on the development of sustainable land projects and film the journey of established community models into a documentary about our shared future.

The present focus on developing powerful AI minds leads us to a power and citizenship enigma that requires our voices to be adequately formed and heard.

If we assume catastrophic scenarios of AI development, then our authorship of our reality and the AI that influences it could disappear. Now is the opportunity to leverage global networking and translating technology to experiment and refine robust dialogue, democracy and cooperation. Within a rigorous system, global priorities and values can emerge. Defining global citizenship and what civic participation looks like in that paradigm is the first question it can start to answer.

I’m approaching this challenge by developing a digital ecosystem of literacy and connectedness, decentralization where possible and connecting the unified approach through living spaces of culture that I then record into the documentary.

The opening, in my words

I’m developing a set of protocols to unify resilient infrastructure standards online and beyond. I’m looking to work on the development of sustainable land projects and film the journey of established community models into a documentary about our shared future.

The present focus on developing powerful AI minds leads us to a power and citizenship enigma that requires our voices to be adequately formed and heard.

If we assume catastrophic scenarios of AI development, then our authorship of our reality and the AI that influences it could disappear. Now is the opportunity to leverage global networking and translating technology to experiment and refine robust dialogue, democracy and cooperation. Within a rigorous system, global priorities and values can emerge. Defining global citizenship and what civic participation looks like in that paradigm is the first question it can start to answer.

I’m approaching this challenge by developing a digital ecosystem of literacy and connectedness, decentralization where possible and connecting the unified approach through living spaces of culture that I then record into the documentary.

The polycrisis, drawn in layers: crises on the surface, a consensus problem beneath them, and a design problem beneath that.A section through the ground under the sky. On the horizon, a skyline of interlocking rings: the polycrisis. In the layer beneath, arrows pointing every which way: a consensus problem, where our voices have to be formed and heard, through dialogue, democracy and cooperation. In the deepest layer, a drawn grid with three foundations joined by one line: shared standards online and beyond, sustainable land projects, and a documentary of community models. Each sends a shaft up to the surface, and the arrows beside each shaft turn to point the same way.shared standards, online and beyondsustainable land projectsa documentary of community modelsnear a foundation, arrows alignThe polycrisiscrises overlapping, on the surfacea consensus problembeneath it: people pullingdifferent waysour voices, formed and heard;dialogue, democracy, cooperationa design problemand therefore: sharedfoundations, joinedThe polycrisis, drawn in layers: crises on the surface, a consensus problem beneath them, and a design problem beneath that.A section through the ground under the sky. On the horizon, a skyline of interlocking rings: the polycrisis. In the layer beneath, arrows pointing every which way: a consensus problem, where our voices have to be formed and heard, through dialogue, democracy and cooperation. In the deepest layer, a drawn grid with three foundations joined by one line: shared standards online and beyond, sustainable land projects, and a documentary of community models. Each sends a shaft up to the surface, and the arrows beside each shaft turn to point the same way.sharedstandardslandprojectsthedocumentaryThe polycrisiscrises overlapping, on the surfacea consensus problembeneath it: people pulling different waysour voices, formed and heard; dialogue,democracy, cooperationa design problemand therefore: shared foundations, joined

The research · the Legible AI lab

Self-correction is competence‑gated. Verification fails where it’s needed most.

A checker helps on problems it can nearly solve itself. Outside that range, most of what it flags is noise, and two models are wrong the same way more than half the time.

The full result →In plain English

The competence gatemodel
0501000%50%100%objections mostly noisemostly realthe gatechecker competence on this task →share of its objections that are real

At 38 competence, 4 of 20 objections are real. most of what it flags here is noise.

An illustrative model of the measured pattern, not the data.

  1. −0.87

    between a checker's competence and how often it wrongly flags correct work. Weak checkers flag 59–67% of right answers; strong ones 2–3.5%.

    n = 7 auditor models95% CI −0.98 to −0.34replicated −0.86 / −0.84 / −0.90

  2. ρ ≈ 0.55

    how often two models give the same wrong answer when both are wrong.

    catch rate tracks 1 − (1 − q)·ρ, corr 0.83

  3. 0.0 → 0.69

    self-repair on a task sampling never solved, where the model had a foothold.

    matched call budgetreplicated

Local open-weight models on graded tasks, and three models from one company on one injection probe. Every limit and every correction is on the research page; the record shows 2 findings withdrawn in full and 0 under correction, of 44 findings on the record at 27 September 2026.

The work

Research programmes

  • Boundary Intelligence

    Simple mechanisms at system boundaries, tested against far larger internal models.

    Synthesis · five negative results

  • Project Reinnman

    Failure-aware tools for AI-assisted mathematics, tested against Robin’s inequality.

    RH open · results labelled by ceiling

Tools

  • Local verification engine

    A local confidence layer that answers with a receipt or refuses with a reason.

    Private alpha · external receipt open

  • Lattice

    A local, OpenAI-compatible gateway with model routing and a containment check.

    Working software

  • lattice-commitsource

    A test-gated wrapper that stops code-repair loops undoing good fixes.

    Alpha · PyPI

  • Governorsource

    Sends each request to the cheapest model tier that can handle it.

    Open source · MIT

Now A second research line is on the record: readable by construction.

Where this work needs other people

Every project is free and noncommercial. These are the places they need someone.

The four openingsfuturisminstitute@gmail.com

Reviewers of the civic half: the Values Commons grant brief is at valuescommons.org/funders.