Skip to main content

Navigation

  • Client impact
  • Writing

AI Solutions

  • AI Systems
  • AI Development

Software teams

  • Design Systems
  • Product Design & Engineering
  • API Platforms
  • View all services for software teams
Get in touch
  • AI Systems

    Model-agnostic AI systems and agents that do real work

    AI Development

    Custom apps and tools built with AI-enabled engineering

  • Design Systems

    Depth of experience that few can match

    Product Design & Engineering

    Experts in cross-functional collaboration

    API Platforms

    Node.js, React, & GraphQL at scale


    View all services for software teams
  • Client impact
  • Writing
  • Contact

Authors

joe prisk avatarjoe prisk avatar

Joe Prisk

Engineer who specialises in building thoughtful, scalable digital products that turn complex problems into simple, human-centred solutions.

Specialties

  • Front-end Engineering,
  • Design Systems
  • article

    Build vs buy: AI code review for teams that have opinions

    There’s a good chance that somewhere in your org, someone is trialling an AI code review tool right now, and at the same time someone else is pasting agent output into PRs under their own name.

    July 2026·6 min read
  • article

    What engineering teams owe for the AI speed boost

    AI-assisted development genuinely accelerated engineering teams, but many organisations stopped at the velocity gains and ignored the accumulating debt. The costs: code nobody fully understands, weakened code review culture, bloated dependency trees, and architectural decisions made without real judgement about tradeoffs...

    May 2026·4 min read
  • article

    Why your design system is the most important input to Claude Code

    Design systems are critical inputs for AI-assisted development tools like Claude Code because they constrain the AI’s micro-decisions around colours, spacing, and components – preventing "drift at speed" and ensuring generated UI remains consistent with your product’s visual language.

    March 2026·5 min read

More posts from other Thinkmill voices

  • AI search explained: indexing, retrieval and reranking

    Organisational knowledge is usually spread across a dozen tools, and keyword search only finds results that match exact words. We run text search, embedding search and reranking on the same dataset, and find out how semantic search with a reranker returns the most relevant results for a query.

    October 2026·12 min read
  • Your agent believes that docs folder

    The agent trusted a committed docs folder of generated artefacts over my instructions. I deleted the lot and lost nothing.

    September 2026·8 min read
  • Are your AI agents making progress or producing work?

    We shipped a working MVP for the Tech Council of Australia in two weeks using an agentic development workflow. Along the way, we learned how easily agents can produce convincing work without moving a product forward, and why engineers still need to decide what progress looks like.

    September 2026·5 min read

Want to keep the conversation going?

Get in touch

Footer

Thinkmill is a member of the Tech Council of Australia(opens in new tab).

AI Solutions

  • AI Systems
  • AI Development

Software teams

  • Design Systems
  • Product Design & Engineering
  • API Platforms
  • View all software teams

Client impact

  • ELMO
  • Tiimely
  • Atlassian
  • Reckon
  • AnywhereWorks
  • View all client impact

Open Source

  • Keystone (opens in new tab)
  • Keystatic (opens in new tab)
  • React Select (opens in new tab)
  • Classnames (opens in new tab)
  • Manypkg (opens in new tab)
  • Changesets (opens in new tab)