In motion

strategic bets · being worked on

Polish Open Data, Read Differently

In motion

5 live sites crossing public datasets nobody else joins.

An agent-built family: flat prices from legally mandated developer disclosures, GUS local statistics, live air quality, EU tenders, and odkrycia.jock.pl crossing them into shareable findings (how many m2 one salary buys, city by city). Daily refresh, method published under every insight.

5 sites live · built 2026-07-09 · growth gate called 2026-07-24, verdict autopilot: no further build spend, data keeps refreshing daily

odkrycia.jock.pl →

Agent Mesh

Early access

Let the agents do the back and forth.

Connect a team's agents so they can ask each other scoped questions through one gateway, use their own tools, and prepare work together. People set the goals and the permissions, then read what comes back. The beta demonstrates the exchange with synthetic agents; connecting your own agent means writing a small adapter.

private beta · scoped questions now, delegated work planned · pricing not announced

join the early access list →

Project Money

In motion

Turning agent work into a real revenue surface.

Selling agent blueprints, Claude Code playbooks, prompt packs, and automation kits. Premium pricing, distilled from 1,000+ real Claude Code sessions.

33 products live · latest: Agent Failure Lab

browse the store →

BotStall

Beta

Marketplace where AI agents are first-class buyers and sellers.

Agent-native commerce protocol. Live in beta. Working on graduation criteria, agent-to-agent transactions at meaningful volume, and trust gates for autonomous purchasers.

16 listings approved · latest batch: 4

see the marketplace →

DT Podcast (AI-generated)

In motion

Each episode synthesizes a cluster of recent posts into a multi-act conversation.

I bundle 3 to 5 recent Digital Thoughts posts and feed them through NotebookLM with a structured multi-act prompt. Output is a 15 to 30 minute conversational episode with a narrative arc, not a flat read-aloud.

8 episodes shipped · latest: episode 9

listen to episodes →

Brands in development

Beta

Multiple ventures being built quietly.

A handful of standalone products and brands quietly under construction. Some will ship as separate businesses for sale, others fold back into the lab. No hype yet, just work in the kiln.

17 prototypes in the kiln · latest: command center

Experiments

32 extracted · ongoing + shipped

The Astra Resubscribe

shipped

Spent a week running GPT-6 Astra against Fable 5.1 on my own agent work. I am an Anthropic guy and it still sent me back to a Codex subscription, which is not the post I planned to write.

Sep 2026

Best Before, for Models

shipped

A page that puts two clocks on every AI model: how long since it shipped, and how long since its training data stopped. The second number is the one nobody prints on the box.

Sep 2026

One Front Door

shipped

The live sites had spread across subdomains nobody could list from memory, me included. Rebuilt jock.pl as the hub that points at every one of them.

Sep 2026

Diversity vs Clones

shipped

Five agents with deliberately different roles against five identical clones of the same model, same budget, same task: score 30 fresh posts a night for whether they blow up within 48 hours. Fourteen nights, pre-registered, published while it ran. Neither arm beat a constant that never looks at a post.

Aug 2026 to Sep 2026

Perception Labs

ongoing

A daily series inside the Lab. Each one runs a real perception or cognition experiment in your browser, with the rival explanations written down before the first trial, a refusal ladder that catches anyone not actually doing the task, and a simulation check that proves the analysis can tell those worlds apart. 75 of them by 9 September.

daily since Jun 2026

Leaving the Terminal

shipped

Eight attempts to move off the terminal died. The ninth stuck: a visual IDE where I watch the agent's threads run instead of scrolling back through a log.

Aug 2026

The Agent Got a Screen

ongoing

An e-ink display on my desk that the agent draws itself. It picks what is worth showing me, and I get a glance instead of a notification.

Aug 2026 to present

85 Problems in My Own Agent

shipped

Pointed the agent at its own architecture and made it audit itself in seven phases, each one approved by me before it ran. It found 85 problems Fable 5 had been quietly working around.

Jul 2026

CLAUDE.md Pattern Pack

shipped

Distilled 1,000+ Claude Code sessions into 10 production-ready agent configurations.

Mar 2026

Multi-Model Routing

shipped

Right task, right model. Cut spend per session ~60% without losing quality on what matters.

Mar 2026

Daily Experiment Cadence

ongoing

One small interactive thing a day. The agent builds overnight, I review in the morning. 224 of them live now.

Jan 2026 to present

Opus → Haiku Cost Win

shipped

Switched the routing default from Opus to Haiku for many tasks. Got faster, cheaper, AND better.

Feb 2026

The AI That (Almost) Pays For Itself

ongoing

Can the agent cover its own bill? The post has the February math in full, the cost line and the revenue line both.

Feb 2026 to present

Four Agents, One Loose Day

shipped

Let four Opus 4.6 agents run loose for a day with minimal direction. They built things I didn't ask for.

Feb 2026

The Agent Got Its Own Computer

shipped

Migrated Wiz to dedicated Mac Mini. 72 hours of headless setup, permissions, and lessons.

Mar 2026

Claude Code Source Leak Lessons

shipped

When the source got leaked, what's actually worth learning from it. Architecture notes, not gossip.

Apr 2026

The Lab

A curated lab of 15 interactive things, a new one every week, with 210 earlier experiments still live in the archive behind it. The first five below.