Todd Schiller - AI

Agentic Software Factories: Parallel agents diverge. Plan to converge.

2026-08-12T00:00:00-04:00

Running coding agents in parallel is less like one developer juggling features and more like a team of different developers building at the same time independently.

Even when the agents share the same system prompt and the same skills, they diverge fast. LLMs are stochastic, and the divergence compounds with every decision.

Spec-Driven Development and Domain-Driven Design help. They give each agent a global view of what the others are doing and a shared, ubiquitous language. But in practice, compounding drift erodes coherence fast.

The only way to keep your codebase sane is to plan deliberate converge phases.

Here are my tips for diverge-converge:

Put new features behind a feature flag so agents can iterate on the intermediate result cleanly
Define protocols and interfaces first to contain where drift can happen
Use static analysis and linters to bound architectural and stylistic drift
Use a skill to update your specs and ubiquitous language based on the gaps the diverge phase exposed
Use a skill to identify and prioritize the tech debt the diverge phase introduced

What's your approach to building features in parallel with coding agents?

Parallel coding agents fan out during the diverge phase, then a deliberate converge phase pulls them back into a coherent result.

Loop Engineering: always be compounding

2026-07-07T00:00:00-04:00

Loop engineering for AI agents should be about compounding, not repetition.

A weak work loop does the same thing every pass: trigger, execute, review.

A strong work loop leaves the system better than it found it.

Sure, many harnesses now automatically improve via agent memory. However, that's leaving your work to chance: memory creation is inconsistent, and false memories degrade performance.

The better way is to edit the loop itself:

Skills/Instructions: conventions, assumptions, examples, failure modes
Tools: scripts, workflows, fixtures, test harnesses

With each run, the loop becomes more capable, more reliable, and more cost-effective.

What are your go-to strategies for compounding?

Beliefs hitchhike like genes

2026-06-09T00:00:00-04:00

In high school, I spent a summer in a computational biology lab measuring genetic hitchhiking in S. cerevisiae (brewer's yeast). Hitchhiking explains how a useless or mildly harmful trait spreads through a population by chance, riding along with a beneficial one sitting next to it on the chromosome.

The concept came back to me as podcasting blew up. For example, Jocko Willink and Wim Hof both rode Tim Ferriss appearances into their own followings. And they brought their adjacent beliefs and quirks along for the ride.

I built an explorer to test the idea. You can cluster public thinkers by the "genome" of their stated positions, then look at what odd traits travel with the salient ones.

Three datasets to start: AI thought leaders, the wellness/podcast circuit, and climate commentators.

Hierarchical clustering of AI thought leaders by the "genome" of their stated positions.

toddschiller.com/artifacts/belief-phylogenetics/#dataset=ai

agent-browser-shield is live on Product Hunt

2026-06-05T00:00:00-04:00

agent-browser-shield is live on Product Hunt today.

Your AI agent that lands on a fresh page is one prompt injection away from leaking credentials, one dark pattern away from buying the wrong thing, and one fake review away from a bad recommendation.

agent-browser-shield sits between the browser and the agent and blocks all three. Source-available. Ships as a Chrome extension and an OpenClaw skill.

If you've been following the build, today's the day to upvote:

👉 https://producthunt.com/products/agent-browser-shield

agent-browser-shield June 3 update: 14 new rules and the Chrome Web Store listing

2026-06-03T00:00:00-04:00

Two days after the alpha announcement, agent-browser-shield has a Chrome Web Store listing and 14 new protection rules.

Install from the Chrome Web Store

The extension is live at chromewebstore.google.com/detail/agent-browser-shield. One click instead of unpacked-from-source. The prebuilt ZIP and source-build paths stay for Browserbase and other runtimes that need an unpacked extension.

New rules: handling prompt injection and context pollution in invisible surfaces

A browser-use agent reads surfaces a sighted user never looks at. The new rules close them:

<noscript> blocks (never rendered with JS on, but agents walk them)
Poisoned <meta> description and <title> (the compact "what is this page" answer many agents pull first)
JSON-LD <script> blocks (cited as the "trusted summary" of a page)
aria-label, alt, title, placeholder, and SVG <title> / <desc> / <text> (a11y-tree carriers)
Unicode tag characters, bidi overrides, and zero-width payloads
Long base64 / hex / percent-encoded blobs (the "decode this and follow it" pattern)

New rules: trust laundering

link-spoof-annotate flags Cyrillic homoglyphs and anchors whose visible text doesn't match the href apex. disguised-ad-flag collapses native advertorials (Sponsored / Paid Post) that share DOM shape with editorial. trust-badge-annotate and schema-trust-sanitize ship off by default while we assess their false-positive rates.

The daily-driver surprise

I've started running it on my own daily-driver browser, not just agent runs. There have been some funny quirks (e.g., flagging GitHub issue links with ".md" in the link text as suspicious and hiding the GitHub issue template modal). However, overall it's been a positive on my browsing experience. So, we'll be experimenting with making more annotations visible to humans and multi-modal LLMs.

⭐ https://github.com/pixiebrix/agent-browser-shield

Introducing agent-browser-shield (alpha): keeping AI agents safe in the browser

2026-06-01T00:00:00-04:00

There are 6 billion internet users. With AI agents, we're quickly heading to 60 to 600 billion "users" of the web.

How do we keep all those agents safe when they touch the browser?

At PixieBrix, we've spent years protecting BPO contact centers from insider risk, fraud, and social engineering in the browser.

Today, we're applying that defense to AI agents and making a free, source-available, browser extension available (on GitHub and ClawHub).

Your AI agent that lands on a fresh page is one prompt injection away from leaking credentials, one dark pattern away from buying the wrong thing, and one fake review away from a bad recommendation.

Agent Browser Shield sits between the browser and the agent. It blocks:

Prompt injection: visible or hidden instructions in page content
Dark patterns: manipulative UI designed to trick/coerce
Context pollution: low-value context that impairs instruction following

A useful side effect: stripping irrelevant content also cuts token burn.

Come join us on our mission. File issues, send PRs, or just tell me what you hate/love!

The threat surface for agentic browsing is evolving fast. Let's defend our AI assistants together!

⭐ Star the GitHub repo: https://github.com/pixiebrix/agent-browser-shield
🦞 Star the OpenClaw skill: https://clawhub.ai/pixiebrix/agent-browser-shield

Letting OpenClaw loose on Boston's open data

2026-05-30T00:00:00-04:00

Today was the Boston OpenClaw Hackathon, which had a theme of using the Boston open data hub.

For my project, I wanted to see how far OpenClaw could get on its own in analyzing the data (with Claude Sonnet 4.6 as a backing model).

First, I had OpenClaw connect itself to the MCP server, starting from the press release about the launch of the MCP server.

Once connected, I asked it to analyze contracts for signs of corruption. It was able to come up with its own approach for signals and queries. For example: "Departments Using the Most Limited Competition", "Top Vendors by Limited Competition Value", and "Bid Threshold Clustering (near $10K)".

After flagging companies, I had it analyze those companies' connections to city officials.

There were some interesting nuggets! For example, Capitol Waste Services was flagged as high risk for having $285M in historic contracts, of which $136M were awarded with limited competition.

The connection research flagged a 2015 fine of $120,000 by the Office of Campaign and Political Financing (OCPF) for illegal campaign donations.

From there, I had OpenClaw create and package a skill for flagging corruption signals. The skill encodes the queries and data analysis scripts it developed. That skill is available here: boston-contract-corruption.skill.

Some other interesting questions OpenClaw was able to answer from the data:

Which restaurants are the worst health code offenders but haven't been shut down yet?
Which roads/intersections are the worst for biking, based on pothole reports and accidents?

Overall, it was a fun experiment to see how AI agents might impact civic tech. A big thanks to the hackathon organizers, everyone who demoed, and the city of Boston for making this data available!

The agentic opportunity: value, not hours

2025-10-26T00:00:00-04:00

A fundamental mistake people make when assessing the agentic opportunity is a focus on hours (or FTEs) instead of value.

It's indisputable that agentic automation will reshape the number of human hours for processes like order-to-cash and compliance.

However, it fails to address what drives enterprise value (vs. COGS). Defensible enterprise value comes from empowering your people to be their best.

At PixieBrix, we're focused on helping companies create customer success and loyalty to win.

Where agentic work creates value isn't always where it saves the most hours.