Jac: Michigan Professor's Python-Like Language for AI Agents

What happened

A computer science professor at the University of Michigan has created Jac, a new programming language modelled on Python that is designed for building AI agents, according to The Register. The report presents Jac as a way for engineering teams to avoid expensive rewrites of existing AI systems.

The Register's summary frames the cost problem as a figure of $165,000 spent on AI rewrites. The language is positioned as a means of bringing together the mix of programming languages that many organisations already run, rather than replacing them wholesale.

The source material available for this briefing is limited to the headline and summary, so details on Jac's syntax, adoption and availability should be checked against the original report.

Why it matters

The development points to a broader shift in how AI agents are built. As organisations move from single chatbots to systems that chain tasks, call tools and act across services, the choice of language and architecture increasingly shapes how quickly those systems can be changed and maintained. A language purpose-built for agents suggests that the tooling layer is becoming a design decision in its own right, not an afterthought.

For leaders in digital transformation, the more immediate question is whether a new language reduces the cost of modernising legacy and polyglot stacks, or simply adds another layer to manage. Any claim about avoided rewrite costs should be tested against an organisation's own systems before it informs budgets.

By the numbers

• $165,000 is the figure cited in The Register's summary as the cost of AI rewrites that Jac is positioned to help avoid.
• 23 October 2026 is the publication date of The Register's report, based on the article URL.

The Renascence take

Most readers will focus on the language itself and miss the operational point underneath it. The real experience question is not which syntax wins, but how much friction your teams and customers absorb every time an AI agent is rebuilt to fit a new stack.

Treat a headline cost saving as a hypothesis, not a business case. Map where your agents currently break or stall, and measure the time spent on rework before you consider any new tooling. The customer impact of a rewrite is usually felt in delayed fixes and inconsistent responses long before it shows up in the budget. A language that promises to reduce rewrites is only worth adopting if it shortens the path from a customer problem to a reliable agent behaviour.