Author(s): Tim Urista | Senior Cloud Engineer

Originally published on Towards AI.

Everyone shipping AI side projects publishes the demo. Almost nobody publishes the ledger. Here’s mine.

This is the complete cost history of Trendvesting, an AI signal-intelligence platform for equities and options that has been running in production since a first commit dated 2024-02-14 and now spans 1,672 commits across a multi-mode Go backend, a Next.js app, a React Native client, and a Python FastAPI consensus service.

After laying out the project’s background, the author argues that the biggest costs of a “small” production AI system aren’t the model tokens but the fixed infrastructure required to keep the system running reliably (e.g., Kubernetes clusters, backups, observability), which makes pricing-page token assumptions misleading. They quantify real spend, show how optimization changes the cost curve, and describe three “taxes” that drove their learning: paying for bad or free data (especially important for options), duplicating work due to missing caching/deduplication, and wasting money on verbose output by not constraining response formats (the input/output price asymmetry makes verbosity a direct charge). They also cover a March migration to cheaper models, measuring signal-loss impact to prove smaller models can work when paired with disciplined risk management (notably, stop-loss execution matters more than model sophistication). Finally, they compute fully loaded cost per signal (about $1.50–$2.00) to emphasize unit economics dominated by fixed costs at small scale, and conclude by encouraging readers to publish their own fully loaded ledgers—calling that “honest” number the rarest thing in production AI.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI