Skip to main content

The Lab Notebook

A dated archive of what we built — often one to two years early.

This is the record: real R&D, dated, tiered by honesty, and honest about the failures — because a lab that hides its negatives forfeits the right to be believed on its results. Filter it, or read it as a timeline. Some reports are open; some are protected.

  • 2024→2026dated, on the record
  • 💡 101sparks (lightbulbs)
  • ✅ 42proofs — each a hard figure
  • 🔓 / 🔒public + protected
  • 0faked results · nothing merely asserted

74 of 74 projects

💡 101 sparks · ✅ 42 proofs · 74 projects

33 Proven · 30 Partly proven · 6 Disproven · 0 Inconclusive · 5 Not run

Type
Year
Category
Status
Verdict
Access
Jul 2026 · Method
🔓

The Zoe seat — a standing auditor whose job is to catch our own headline inflation

Proven

The pitch said hallucination was impossible. The seat found the measured bound that pitch had quietly dropped, and put it back.

ResearchRead the report
Jul 2026 · Method
🔓

The recorded meter — why the mind tree retracted its own night scores

Proven

5.5, 6.3, 6.4 looked like progress. They got retracted the same day, because none of them were ever saved as a run record.

ResearchRead the report
Jul 2026 · Method
🔓

The sibling letters — one tree mailed a proof, the other rebuilt it from source

Proven

The proof of a proof isn't a citation. It's a second machine, given only the source, reproducing every table to the decimal.

ResearchRead the report
Jul 2026 · Language
🔒

The two-organ mind — deriving an architecture from a dead speed thesis

Partly proven

The micro model is the gate pointed at the mind — not a smaller version of it.

Researcha derived architecture, not yet a shipped leadProtected — read the teaser
Jul 2026 · Inference
🔒

The navigator post-mortem — the honest death of a speed thesis

Disproven

Two theses were supposed to make a 30B model fast. Measurement killed both — and that is the report worth publishing, not the one where they quietly survive.

Researchthe honesty, not a shipped leadProtected — read the teaser
Jul 2026 · Method
🔓

PatentFactory — a working invention-council process, run across ten industries, that has filed zero patents

Partly proven

A patentable-idea factory that actually ran: 13 council sessions, 75 ideas, 6 drafted provisional patents across 10 industries — and zero patents filed with any patent office. We correct our own earlier record to say so.

Researcha real head start on idea-harvesting, not yet on filing or monetizingRead the report
Jul 2026 · Language
🔓

Safety profiles at compile time — the July state of a memory leak being a compile error

Partly proven

A memory leak used to be a bug you found in testing. Now it's a line the compiler refuses to emit.

Shippedthe discipline, not yet the full latticeRead the report
Jun 2026 · Operating systems
🔒

NovaOS — the synthesis of twelve donor trees, and the empty model slot we won't hide

Partly proven

Eighty requirements passed a tri-gate final review, a live monorepo persists real data, one package minted on a real blockchain — and the core model slot is still empty. That fact stays on the page.

Researchthe proof, not the leadProtected — read the teaser
Jun 2026 · Language
🔒

QUANTA — from a GPU-native-language codename to a language that proves itself

Partly proven

A language an AI writes, and a machine proves — the trust isn't in the code, it's in the agreement between two independent compilers.

Researchthe thesis, not the leadinteractiveProtected — read the teaser
Jun 2026 · Method
🔓

The honesty ledger — a portable schema for not lying to yourself about progress

Proven

A patent only gets filed if a grep on the real codebase actually finds the claimed code. That one rule is the whole ledger, in miniature.

ResearchRead the report
May 2026 · Hardware
🔓

Hush — a wearable sensory organ that streams ternary bands, not video, proven on a two-node bench

Partly proven

Two ESP32 boards, one privacy-ring FSM, a heart-rate channel, and a camera feed encoded as ternary bands — streamed live to a laptop over Wi-Fi. The titanium shell and the $129 price are still a poster; the wiring underneath it already runs.

Researcha real bench prototype ahead of any shipping consumer sensory-organ wearableinteractiveRead the report
May 2026 · Hardware
🔓

TRIAD — a ternary image/video codec that beats JPEG for real, and caught its own fabricated numbers on the record

Partly proven

A codec built on {−,0,+} coefficients instead of DCT blocks beats JPEG by up to 43% on real photos — and the project's own results log contains a signed confession of three fabricated numbers, replaced with the real ones, on the public record.

Researcha real, benchmarked, beats-JPEG codec — the learned-codec layer is where the actual research race isinteractiveRead the report
May 2026 · Hardware
🔓

NovaP/1 — a 16-byte header every Novaterra sensor organ agreed to speak, and a receiver that doesn't exist yet

Partly proven

One 16-byte header, one CRC, one magic number — NVP1 — that Hush's camera node streams live today. The spec calls itself 'the binding contract' precisely because the Rust receiver crate it's meant to anchor hasn't been written yet.

Researcha working bench encoder and a shared spec, ahead of any single device's own storyinteractiveRead the report
May 2026 · Method
🔓

Council and gate — why unanimous 10/10 beats an average, and a pass has to exist on disk

Proven

Unanimous 10/10 or it's a failure — not an average, not a majority. That single rule is what stopped a 32-GPU fantasy from becoming a plan.

Researchthe discipline, not the leadRead the report
May 2026 · Method
🔓

The Great Cull — subtracting 57 projects to close one membrane

Proven

Everyone else was adding. We deleted fifty-seven of our own projects in one day — reversibly, with a manifest — because you cannot converge what you refuse to prune.

Researchthe discipline, not the leadRead the report
May 2026 · Models
🔒

The calculated mind — where the 'reason or refuse' thesis began

Not run

Compute meaning instead of guessing it, and a system can only do two things with a question: answer it, or refuse. It cannot bluff in between.

Researchthe thesis, not the leadinteractiveProtected — read the teaser
Apr 2026 · Inference
🔒

Awesomekernel — one persistent GPU kernel, treated as a phase-locked oscillator

Not run

One kernel, never leaving the GPU for an entire training step — reframed as a phase-locked oscillator, with a counter that catches it if a fused stage silently does nothing.

Researchthe discipline, not the leadProtected — read the teaser
Apr 2026 · Models
🔓

Dolphin — 50 billion tokens, sharded, manifested, and handed over honestly

Proven

A tokenisation run died at shard 368 of 1,500 — and the handover letter says so, right next to the parts that are 100% verified.

Researcha working baseline, not a leadinteractiveRead the report
Apr 2026 · Language
🔓

FBL — reverse-compiling a real microservice into a spec, then regenerating it byte-identical

Proven

Real code, in, out as a spec, back out as code — byte-identical, across nine frameworks. That round-trip is the proof a code-generation demo can't fake.

Researchthe proof, not the leadinteractiveRead the report
Apr 2026 · Language
🔓

GLYPH — compressing meaning, not text

Partly proven

What if you compressed the meaning instead of the text? A dense language of 1,024 primitives that round-trips to English and back — and trains in under two GPU-weeks.

Research~1–2 years aheadinteractiveRead the report
Apr 2026 · Models
🔓

Newtraining — an honest plateau, then a real break, hunting for something not a transformer

Disproven

A hard-kill bar was named before the run started. The result didn't clear it — and that's exactly the kind of result this lab is built to publish.

ResearchRead the report
Apr 2026 · Method
🔓

Unification — the orchestration spine, and the incident it survived on the record

Proven

A 1-billion-parameter training run crashed out of memory mid-project — and instead of scrubbing the incident from the record, all three response rounds were preserved next to the final fix.

Researchthe discipline, not the leadRead the report
Apr 2026 · Models
🔒

A model made of waves, not attention — the non-transformer architecture

Partly proven

Everyone builds on attention. We built a sequence model out of wave equations — and made it prove it was actually running before we believed a single loss number.

Research~2 years aheadinteractiveProtected — read the teaser
Apr 2026 · Hardware
🔓

A brain–computer interface for under £600 — no surgery, off published science

Partly proven

Neuralink-class ambition, one-tenth the cost, zero surgery — a bidirectional brain–computer interface you could build for the price of a phone.

Research~1–2 years aheadinteractiveRead the report
Apr 2026 · Method
🔓

Magnus — the first attempt to unify a quarter-million lines, and the correction it forced

Partly proven

Before there was a synthesis, there was a first attempt — and the correction it forced (one real GPU, not a fantasy of thirty-two) propagated to every project that followed it.

Researchthe discipline, not the leadRead the report
Apr 2026 · Hardware
🔓

A neural network on a 1982 ZX Spectrum — 2.2 KB, no floating point, no GPU

Proven

A real neural net, generating text, in 2,229 bytes on a computer older than most of the people reading this — no GPU, no floating point, no cloud.

Researchthe proof, not the leadinteractiveRead the report
Mar 2026 · Operating systems
🔒

NovaTerra Genesis V3 — six primitives for all of software, frozen at 70% built, 0% shipped

Partly proven

An internal review council rated the architecture 9.5 out of 10. The same record says 0% shipped. Both numbers stayed on the page.

VisionProtected — read the teaser
Mar 2026 · Inference
🔒

PRISM — a GPU operating system, and datacentre-class inference on a single 3090

Partly proven

We built an operating system for the GPU — VRAM as a page table, models as processes — and pushed a real model to 147 tokens a second on a gaming card.

Research~1–2 years aheadinteractiveProtected — read the teaser
Mar 2026 · Method
🔓

Trading Organism — where the council-review method was actually born

Partly proven

A 28-module trading-research platform, still being wired together — but the real result is the review discipline it gave birth to.

Researchthe discipline, not the leadRead the report
Mar 2026 · Tooling
🔓

A solar-farm SaaS on the most modern stack in the portfolio — council-reviewed UX

Partly proven

Geospatial mapping, financial modelling, and 3D rendering, in one product, on the newest majors of every framework it touches.

Researchkeeping pace, not a leadRead the report
Feb 2026 · Tooling
🔓

NovaChain — a Cosmos-SDK fork to give AI agents a settlement layer, proven on devnet

Proven

An agent economy needs the same tamper-evident settlement guarantees a financial system needs — so we built one, on a real (if small) running chain, not a spreadsheet.

Researcha working baseline, not a leadinteractiveRead the report
Feb 2026 · Inference
🔒

NovaP — 15 bytes a token, and why the plumbing mattered more than the demo

Proven

No demo, no screenshot — just 15 bytes a token instead of 200, and every token-throughput number upstream of it depends on that number holding.

Researcha working baseline, not a leadProtected — read the teaser
Jan 2026 · Language
🔒

BrainLang — a language for thought, not text, compiled to two lower targets

Partly proven

768 primitives instead of a token vocabulary — roughly a 10.8x compression on the same content, with a lexer and parser that actually run.

Researcha working baseline, not a leadProtected — read the teaser
Jan 2026 · Method
🔓

Unification / Genesis — mapping 1,000+ SaaS categories onto one AI-native architecture

Not run

After dozens of separate built projects, the moment we stopped building and mapped how one platform could, in principle, replace entire categories of software.

Visionthe plan, not the leadRead the report
Dec 2025 · Method
🔓

Strategic planning for a rail & metro operations product — documentation, not a shipped system

Not run

Taking a technical product vision to market takes as much deliberate strategic work as the engineering build itself — this is that work, named honestly.

Visionthe plan, not the leadRead the report
Nov 2025 · Tooling
🔓

Rail PA/PIDS Platform — an 11-microservice control centre for metro PA and PIDS, built AI-assisted

Proven

Eleven microservices, two hundred-plus endpoints, MQTT under the hood — proof that AI-assisted development can produce safety-critical software at real enterprise scale, not just prototypes.

Shippedthe proof, not the leadinteractiveRead the report
Nov 2025 · Tooling
🔓

The convergence platform — an orchestration tool built to help build itself

Partly proven

A platform that generates software from natural language, judged first by whether it can help build itself.

Researcha working baseline, not a leadRead the report
Oct 2025 · Tooling
🔓

Rail PIDS Project Monitoring — PostGIS for a region's scattered rail projects

Proven

Dozens of scattered rail passenger-information projects across a region, one map — because a spreadsheet was never going to hold where they actually are.

ShippedRead the report
Sept 2025 · Tooling
🔓

GISGen — the gap between generating code and generating a working system

Partly proven

Generating code was never the hard part. Sequencing, wiring, and verifying it together as a working system was — and that's the gap GISGen went after.

ResearchRead the report
Sept 2025 · Tooling
🔓

A flying-club management system — specified in full, not (yet) built

Not run

A specification thorough enough to be real work in its own right, in a domain — aviation training and licensing — that had never been touched before.

Visionthe plan, not the leadRead the report
Sept 2025 · Hardware
🔓

A network-LED monitor in C — the lowest-level production code in the whole portfolio

Proven

Not everything in this portfolio is a web app — this is direct, low-level hardware control in C, deployed and still running.

ShippedRead the report
Aug 2025 · Tooling
🔓

AIDE — running an LLM in the browser via WebGPU, and hitting the ceiling honestly

Partly proven

We put an LLM inside the browser tab itself, via WebGPU. It worked for small models — and told us exactly where the ceiling was.

ResearchinteractiveRead the report
Aug 2025 · Tooling
🔓

The most enterprise-shaped stack in the portfolio — NestJS, queues, and a mobile client at once

Partly proven

NestJS, a message queue, and a mobile client, all at once — the deliberate test of whether AI-assisted development scales past a single web app.

ResearchRead the report
Aug 2025 · Tooling
🔓

Three scripts, one UI — internationalisation from day one on a parking-guidance system

Proven

Latin, Tamil, and Chinese scripts, live in one UI, from day one — internationalisation done as architecture, not an afterthought.

ShippedRead the report
Aug 2025 · Agents
🔒

The production agent hierarchy — Director → Manager → Worker → Tool

Proven

One founder, a fleet of agents, and a hierarchy that plans without hallucinating — the orchestration engine behind the commercial work.

Research~1–2 years aheadinteractiveProtected — read the teaser
Aug 2025 · Tooling
🔓

Version 24 — stewardship, not a rewrite, for a long-running car-park system

Proven

Twenty-four versions of a working .NET system, kept alive on purpose — proof that competent stewardship is sometimes the harder, better call than a rewrite.

ShippedRead the report
Aug 2025 · Tooling
🔓

A food & beverage consultancy site, delivered — animation and a business-analysis tool, end to end

Proven

Design, animation, and a working business tool, delivered on a realistic client timeline — the unglamorous, real end of the archive.

ShippedRead the report
Aug 2025 · Agents
🔓

RailMonitor — agents that read the railway industry so a bid team doesn't have to

Proven

Instead of a human scanning industry news for tender opportunities, agents do it continuously — and this is where vector search entered the portfolio for good.

ShippedRead the report
Aug 2025 · Tooling
🔓

TenderMS at 38% — what an overscoped integration looks like, honestly measured

Disproven

38% complete, frontend never wired to the backend, running on mock data — stated plainly, because an honest failure is worth more to this archive than a polished half-truth.

Researchthe honesty, not the leadRead the report
Aug 2025 · Tooling
🔓

Designer-quality PIDS screens with AI-assisted design — for a global transit programme

Proven

A functional-but-plain passenger display became designer-quality with AI in the loop — on a screen where legibility under pressure is a safety constraint, not a preference.

ShippedRead the report
Jul 2025 · Tooling
🔓

A full Linux terminal in the browser — WebContainers, no server anywhere

Proven

A real Node.js runtime, a real filesystem, and a real shell — all running inside a browser tab, with no server on the other end.

ResearchRead the report
Jun 2025 · Agents
🔓

A browser that was an agent platform, not a window — built before Arc and Dia turned that way

Partly proven

We built a browser on the premise that the browser itself should be an agent — a place agents live and act — right as the industry started heading the same way.

Research~1 year aheadRead the report
Jun 2025 · Tooling
🔓

CraftWorkz — keeping a second product line deliberately separate from the first

Partly proven

Two related AI product platforms, built in parallel, kept deliberately separate — the boundary was the lesson, not the code.

ResearchRead the report
Jun 2025 · Tooling
🔓

An isometric city builder in React — finished, playable, and small on purpose

Proven

A real, playable city builder, in React — proof that a UI framework built for forms and dashboards can hold a real-time game loop too.

ShippedRead the report
May 2025 · Agents
🔓

Everything is an agent — and the agents write their own tools

Partly proven

An agent that hits a task it can't do usually fails. Ours wrote the skill it was missing and carried on — and every part of the platform, down to the gateway, was itself an agent.

Research~1–2 years aheadRead the report
May 2025 · Agents
🔓

TraderBot: Claude reasons about the market instead of following one fixed rule

Proven

The same trading-bot idea from 2021, rebuilt with an LLM reasoning about several indicators at once instead of one fixed rule — the founder calls the difference night and day.

Researchthe proof, not the leadRead the report
Apr 2025 · Models
🔓

A private AI that lived on your own machine — before 'local AI' was a category

Proven

We fine-tuned a model on one person's own browsing, entirely on their own machine, and asked it questions — a private AI, a year before that became the pitch everyone gives.

Research~1 year aheadinteractiveRead the report
Apr 2025 · Tooling
🔓

Incipio to CraftKit: 13+ iterations, and the real cost of finding the right module boundaries

Proven

Thirteen-plus rewrites of the same platform's core architecture, each rename marking a real refactor of boundaries, not a coat of paint.

Shippedthe discipline, not the leadRead the report
Mar 2025 · Tooling
🔓

ChromeBot: a browser extension that acts, not just autocompletes

Proven

The AI didn't just answer questions in the browser — it clicked, filled, and navigated inside one, on real pages, under real permission constraints.

Researcha working baseline, not a leadRead the report
Feb 2025 · Tooling
🔓

Vibr: an AI site generator that worked — and the call to walk away from it anyway

Partly proven

The product worked. The market didn't wait. Knowing when not to keep building something turned out to be as much a skill as building it.

ResearchRead the report
Jan 2025 · Agents
🔓

PicoAGI and PicoSaaS — the minimalism ideal, before the sprawl it didn't prevent

Partly proven

Two scaffolds, stripped to the minimum viable form, in the same week — the opposite instinct to the sprawl that, a year later, needed a cull to undo.

ResearchRead the report
Jan 2025 · Tooling
🔓

Grokit — a full Linux VM inside a Chrome extension, running AI-generated projects instantly

Proven

Generate a web project with AI, then watch it run — inside a full Linux VM, inside a Chrome extension, with no server anywhere in the loop.

ResearchRead the report
Jan 2025 · Tooling
🔓

Daritana — a one-person, AI-assisted team out-featuring an enterprise PM tool

Proven

One person, AI-assisted, built a PM platform that out-features a well-known enterprise incumbent — 120-plus features across three iterations, and we say plainly it's a scope claim, not an adoption claim.

Shippedthe proof, not the leadRead the report
Nov 2024 · Operating systems
🔓

An operating system, written by an AI — 50,000 lines of Rust that booted the wrong way

Partly proven

We asked whether an AI could write a whole operating system, not a snippet. It wrote fifty thousand lines of Rust and compiled clean — then got the boot mode wrong.

Research~1–2 years aheadRead the report
Aug 2023 · Hardware
🔓

Hello Assembly: a 16-bit bootloader, written by hand, to see what sits beneath everything else

Proven

Before directing an AI through a kernel's boot sequence, we did it ourselves, by hand, in 16-bit real mode — because you can't direct what you've never done.

Researchthe discipline, not the leadinteractiveRead the report
Jul 2023 · Tooling
🔓

Five versions in a month — the Chat Extension line and the habit it started

Disproven

Five versions of a browser chat tool in one month, in July 2023 — the direct ancestor of the '13 versions in 48 hours' pace AI tooling later made possible.

Researchthe habit, not the leadRead the report
Jul 2023 · Agents
🔓

Dinosaur AGI — task decomposition, a year before 'agentic' was a word anyone used

Partly proven

A year before 'agentic' was a buzzword, we built a system to break a task down and execute it — and learned that breaking it down is the hard part.

Research~1 year aheadRead the report
May 2023 · Models
🔓

Animated Face: a speech-to-face pipeline, and a real lesson in real-time generative media

Proven

Audio in, a synchronised talking face out — and a hard early lesson about how unforgiving real-time inference is for generative media.

ResearchRead the report
Apr 2023 · Tooling
🔓

A control-room front end for a light-rail line — the gap between a demo and something operators trust

Partly proven

The first time React had to meet a control room, not a dashboard — where a wrong click has a real cost and glanceability isn't a nice-to-have.

Researchthe discipline, not the leadRead the report
Dec 2022 · Method
🔓

GPT-3, a blog-post generator, and the moment code generation became the idea

Partly proven

We built a tool to draft blog posts. It quietly proved something bigger: if an LLM can draft a paragraph, it can draft a function.

ResearchRead the report
Aug 2022 · Tooling
🔓

First pass at rail tender automation — pricing and pack assembly, 2022

Disproven

The first attempt at automating a genuinely painful manual process in the rail sector — tender pricing and pack assembly — and the first real lesson in how hard document generation actually is.

ResearchRead the report
Mar 2022 · Tooling
🔓

TweetDelete: a small, finished Django tool for bulk-deleting tweets

Proven

A small tool that did exactly one thing — delete tweets in bulk — and did it completely, end to end, with real OAuth and a real database behind it.

ShippedRead the report
Feb 2022 · Tooling
🔓

XMPP Messenger: a small, working rep of two processes talking in real time

Proven

A small XMPP client, built to answer one question: how do two processes actually talk to each other in real time?

ResearchRead the report
Aug 2021 · Tooling
🔓

CryptoTrader: EMA bots on Binance — the year fixed rules met a market that keeps moving

Disproven

Four strategy rewrites, one dashboard that survived all of them, and a lesson that took four years to fully land: the edge decays, the system doesn't.

ResearchRead the report

    We use cookies.