Rough Cuts of Tomorrow

I tend to treat knowledge work like open-source: ship small, ship early, learn in public.

Last year I spent a bit of time prototyping futures-intelligence setups - small, connected systems that use AI to support human sense-making. Not to replace the hard graft of synthesis, but to try and raise the base level from which my thinking starts - helping me see things from different perspectives and visualise shifts when the world swerves. Whilst I’ve parked these for now, I thought I’d share them here just for fun.

Here’s the loop:

• Scanning for weak signals across domains and regions

• Mapping fracture & diffractive zones where pressure and possibility meet

• Clustering emergent patterns (H1/H2/H3) and test their coherence

• Translate patterns into ideas, experiments, and hypotheses

• Integrating what I learn back into my own thinking

• Posting so others can reuse (or critique) the moves

The approach is less “big fancy report” and more living system: tags that evolve, graphs that reveal relationships, and lightweight prompts that experiment with visualising novelty, impact, or plausibility. The aim isn’t prediction; it’s explicit pattern awareness or option-value.

Building in public means sharing the half-right ideas or builds, the misclassifications the agent made, and the prompts that actually moved the needle. It means keeping tags, graphs, and scoring rubrics forkable - so I can compare notes with others.

What I’m learning is that transparency compounds. Regular signal drops create a trace. Open notebooks invite better questions. Quick-and-dirty visual workflows beat glossy trend decks for real feedback and useful learning so I’m going to make a big effort to take the time to interrogate and reflect as I go.

Two imperfect builds from last year I’m sharing here

In advertising we used to call these early creative reviews a “tissue session”—nothing is precious; everything’s testable. Same energy here.

These are just two I spun up over a weekend last year to see what it feels like to see the signals visually surfaced rather than stuck in a database. They’re not “right,” or even close, but they already suggest some relationships are worth mapping (and others less so), and give me a better base for the next iteration.

Build # 1

Build # 2

To say I’m a massive Airtable fan for early build / info architecture planning would be an understatement. I find it a really useful tool to think about what I’m trying to achieve step by step, and to make the process really explicit so I can interrogate / refine it before shifting to a database like Supabase if I’m wanting to try and build a custom front end. . although I’m fast approaching the cost ceiling of what I can trial with in-sheet airtable agent workflows. For those who are interested, Airtable launched a new platform a couple of months ago - Hyperagent which I’m going to play around with next.

What Else

Like everyone since being introduced to Wisprflow I’m a little bit obsessed; and also playing around with Claude cowork to see how it could be helpful.

If you’re working on horizon scanning or innovation-for-good somewhere in a space leaning toward better human futures, especially in development - I’m all ears. Tell me what would be useful.

This isn’t a product play. It’s open research infrastructure - meant to be reused, tested, forked, broken, and rebuilt. Whatever helps.

More soon.