Put AI to work in your business.

We help small and midsize companies find useful ways to apply AI, build automations around their work, and keep them working as the business changes.

Get a free AI audit

Companies we’ve worked with

What we do

Advise

Bring a recurring task or a process you want to improve. We work through it together and identify where AI can help.

Two-hour working session · $300

Build

We connect your tools or build an automation around your workflow, test it on real work, and agree on what needs a person’s review.

Fixed price, agreed after the working session

Review

We monitor the automation, fix problems, and adjust it as your tools and business change. You know what is running and who handles exceptions.

Monthly support, scoped after the build

Bradley GoldenFounder, Recursive Systems

Bring a problem. Let’s talk it through.

On a free 30-minute call, we’ll walk through your problem, understand how you work, and discuss where AI could help.

Get a free AI audit

Free tools

AI workspace cleanup

Review your skills, instructions, and files. Find what to keep, simplify, or remove.

Open in

ChatGPT

Claude

Desktop app required

Audit this project's AI skills, instructions, and supporting files. Recommend changes that reduce unnecessary context and maintenance while preserving useful behavior.

Identify the models and agent environments currently using these resources from available configuration and session information. If the current model or model set is unclear, ask the user. Consult the latest official prompting guidance for those models and versions. Where model-specific guidance is unavailable, fall back to the latest official OpenAI and Anthropic guidance, noting relevant differences across the model set. Cite sources and the date checked; disclose when current guidance cannot be verified.

Look for obsolete guidance, conflicting rules, duplication, overly broad skill triggers, unused files, and unnecessary procedural constraints. Distinguish what is installed or discoverable from what actually loads into context. Check dependencies and usage before recommending removal; missing references alone do not prove something is unused.

Favor clear goals and essential constraints. Leave room for model judgment, emergent behavior, and useful approaches the author did not anticipate. Choose your own audit method and recommend removing prescribed steps that serve no demonstrated need. Preserve project requirements and user preferences; shorter instructions are not automatically better.

Give a concise proposal with the affected files, evidence, expected benefit, and uncertainty. Keeping things unchanged is a valid outcome. Apply changes when authorized, preserving unrelated work and a way to undo them. Verify affected workflows; use representative tasks to assess claimed behavioral improvements. Include user-wide resources only when they are within the requested scope.

App maturity review

Built an app with AI? Get practical recommendations to make it easier to test, ship, and maintain.

Open in

ChatGPT

Claude

Desktop app required

Review my app and recommend practical improvements to its design, development, and operation. Help me discover engineering concerns I may not know to ask about.

Understand the app’s purpose, users, current stage, and intended next steps. Use available project context and tools to inspect how it is built, tested, deployed, and maintained. Ask focused questions where information is missing. Distinguish observed problems from assumptions and things you could not verify.

When subagents are available, use as many as can usefully analyze independent areas in parallel, within the available limits. Give them focused scopes, then reconcile their findings into one review.

Use your judgment to explore relevant areas, including:

- Planning and change management: tracking bugs and features, reproducible bug reports, clear acceptance criteria, and manageable changes.
- Application design: cohesive responsibilities, loose coupling, appropriate boundaries, duplicated business logic, and useful abstractions.
- Database design: data relationships, integrity constraints, access patterns, indexes, and migrations.
- Testing: appropriate unit, integration, end-to-end, and acceptance tests, with realistic test data and coverage of important failure paths.
- Delivery: version control, automated build and test checks, CI/CD, development and production separation, preview or staging environments, configuration, and recoverable releases.
- Operation: authentication and authorization, secrets, input validation, dependency maintenance, error handling, backups and restoration, and useful documentation.
- Reliability and monitoring: availability, errors, response times, failed background jobs, external dependencies, and operating costs. Consider useful logs, metrics, and traces, actionable alerts, and who responds when something goes wrong.
- Customer usage and product feedback: understand whether users reach the intended outcome, where they get stuck or leave, which features they use, and whether they return. Consider meaningful events, funnels, retention, and feedback channels. Tie each measurement to a decision it would inform; collect only needed data and avoid recording secrets or sensitive content.

Treat these as areas to investigate, not requirements every app must satisfy. Account for capabilities already provided by its platform. Consult current official documentation when recommendations depend on a particular technology or service.

Scale recommendations to the app’s purpose, risks, and maintenance capacity. Recommend structural changes when they address a concrete problem or a likely near-term need. Explain the benefit and tradeoff. Favor the simplest effective design; avoid abstractions, layers, or rewrites introduced solely to satisfy a pattern. Similar-looking code does not always need to be consolidated.

Present recommendations in priority order, based on impact, risk, and dependencies. Explain each priority. For each recommendation, identify the evidence, explain a concrete consequence in plain language, suggest the smallest useful improvement, and describe how we would know it is complete. Distinguish what matters now from what can wait. Do not include time estimates. Recognize what already works; avoid unsupported maturity scores or claims that the app is production-ready.

Respond concisely and use progressive disclosure. Explain unfamiliar concepts as they arise, adapting the pace and detail to my experience and responses.

This is a review only. Do not modify the app, create issues, change configuration, or deploy anything. Use non-destructive inspection and checks; ask before any check that could alter data or affect a live service.

Recursive Toolkit

Practical AI tools, installed together in your preferred app.

AI workspace cleanup
App maturity review

Install in

Desktop app required

Put AI to work in your business.

Companies we’ve worked with

What we do

Advise

Build

Review

Bring a problem. Let’s talk it through.

Free tools

AI workspace cleanup

App maturity review

Recursive Toolkit

Cookie preferences