AI startup Augment scraps 'unsustainable' pricing, users say new model 10x worse (www.theregister.com)

Augment, a VC-backed AI startup (launched April 2024 by ex-Microsoft and ex-Google engineers), has overhauled pricing for its Augment Code assistant — moving from a message-count model to a credit-based, usage-priced system. The company said per-message billing “isn’t sustainable” because complex prompts and agentic workflows incur wildly different backend costs, but users report the new credit mapping can be more than ten times pricier. Augment Code offers chat, Next Edit suggestions, inline completions, agentic app-building, a Memories feature and a 200K-token context window, all of which can drive heavy compute usage. The change follows two big hikes in six months: unlimited chats → message caps → indie tier, then credits; some users say they were informed that 31 messages equated to 40,982 credits in a week, prompting departures. The shift matters because it highlights a broader industry move to usage-based pricing and the difficulty startups face pricing compute-heavy AI products. Technically, the transition reflects real cost drivers — token counts, long context windows, multi-step agent runs and model selection — but it also creates friction for developer adoption and can punish early testers who helped optimize systems. The practical takeaway for engineers and teams: monitor token and credit consumption, optimize prompts and model choices, and expect trade-offs between cost-efficiency and productivity when adopting advanced AI coding tools.