Do not use AI when ordinary tools are enough
File search, Git diffs, tests, formatters, parsers, LSPs, and static analysis often produce faster and more precise results than asking a large model.
Do more with less AI.
ChatCode exists to help users and AI platforms use resources more efficiently while making work faster, more accurate, and easier to control. We do not optimize for using more AI. We optimize for the value created by each use of AI.
Do not maximize AI consumption. Maximize the value created by AI.Local when appropriate · Cloud when valuable · Frontier when truly necessary.
Powerful AI is a valuable resource, not a quota pool to drain.
Advanced AI models require large-scale infrastructure, hardware, energy, research, and operations. Economics vary by provider and plan, so we do not assume every request loses money or has the same cost. One principle remains clear: frontier intelligence is a valuable resource and should be used where it makes a real difference.
Move appropriate work local
Log summarization, error classification, context filtering, and drafting can often be handled by an appropriate local model, reducing latency and keeping more data on-device.
Reserve powerful AI for hard reasoning
Architecture, complex bugs, multi-dimensional tradeoffs, important reviews, and ambiguous problems are where stronger models earn their cost.
Prepare context before calling AI
ChatCode can use Project Brain, tools, and smaller models to filter data, build dependency maps, run checks, and give stronger AI exactly what requires reasoning.
Efficiency must not come at the expense of quality.
The goal is not to reduce AI calls at any cost. If a task needs a strong model, ChatCode should escalate immediately. What should be reduced is unnecessary computation: rereading known information, sending irrelevant context, using oversized models for deterministic work, or repeating analysis that can be cached and verified.
Use less. Achieve more. Respect compute. Respect intelligence.
EfficiencyChoose the smallest engine capable of completing the task at the required quality.
PrecisionBetter context, the right tools, and explicit verification improve precision.
ResponsibilityNo artificial consumption, no quota draining, and no business model that turns ChatCode's benefit into ecosystem harm.
A good product should align incentives across the ecosystem.
Users want to save money, time, and quota. AI providers need efficient infrastructure use so they can keep investing in better models. Local models need practical applications. ChatCode needs to create enough value to grow for the long term. We believe these goals do not need to conflict.
ChatCode aims to make the AI subscription a user already pays for more useful in real work while avoiding cloud calls for tasks that do not need frontier intelligence. Done well, users get more value and platforms receive better-prepared, more purposeful requests.
Về môi trường: ChatCode không mặc định tuyên bố “local luôn xanh hơn cloud”. Hiệu quả năng lượng thực tế phụ thuộc model, phần cứng, datacenter và workload. Cam kết đúng của chúng tôi là giảm tính toán không cần thiết, đo lường trước khi tuyên bố và ưu tiên kiến trúc dùng tài nguyên có mục đích.