What happened
The New Stack profiles Uvi, a company aiming to stop AI coding agents from having to relearn the same background again and again. The starting point in the piece is a familiar one: each time an agent has to trawl Slack threads, Jira tickets and old pull requests to piece together why something was built a certain way, that effort is repeated and paid for. We only had the headline and summary to work from, so we won’t describe how Uvi’s product works.
Why it matters
If your team uses coding agents, every session probably starts with the agent gathering context. That means reading a lot of text, and with most hosted models you pay for what gets read. The cost is easy to miss because it shows up as a general model bill, not as a line item for “rediscovering decisions”.
There is a second cost. When the reasoning behind a decision lives in chat threads and closed tickets, the agent may reach a different conclusion each time, or miss the point altogether. New engineers have the same problem, so this is partly a documentation issue that agents have made more visible.
For a company without a platform team, nobody usually owns this. Usage grows, the bill grows, and it is unclear whether the spend is buying useful work or repeated reading.
What to do this week
- Look at your AI tooling invoices and usage reports. Find out whether you can see spend per team, per repository or per tool, and if not, note that as a gap.
- Ask two or three developers where their agents spend time gathering context. Which repositories or topics cause the most back-and-forth?
- Write down the decisions that live only in chat. A short decision record or README section in the repository is cheaper than an agent reading a long thread every time.
- Check what any tool you are evaluating stores about your Slack, Jira and code, who can access it, and how it is deleted. Treat that as a data question, not just a convenience feature.
- Set a rough monthly budget or alert for agent usage, so a change in behaviour shows up early.
Need a hand?
Our Cloud Cost Optimisation work covers this kind of spend, from cloud infrastructure to the newer line items such as AI tooling, and helps you see what is worth paying for. If you would like a second opinion on where your costs are going, the first call is free and it is with an engineer, not a salesperson.
Related service: Cloud Cost Optimisation