AI assistants and search decision
ChatGPT
A competent developer can reproduce the core chat assistant (UI + model relay + history) in about a week and run it affordably; proprietary extensions, plugins, and hosted model scale are not replicated.
Visit website↗Open-source builds that already do this
Every project below is open source and already does this job today. Fork one, self-host it, or take the parts you need — the build prompt further down assumes an empty file, and this is the shortcut past that. Licences differ; check the one on each card before you ship.
What a replacement has to do
- A user types a message in a chat UI, the backend forwards it to a model API, receives the model response, stores the exchange in history, and returns the reply to the UI.
What it still won’t have
- Proprietary plugins ecosystem
- Built-in image generation and voice features as deployed by the vendor
- Hosted model infra and any proprietary model improvements
- Deep integrations and platform polish (saved chats tied to vendor account, account-level personalization)
What remains hard
- Product polish and ongoing maintenance
First-year cost
Keep paying
Paying is—cheaper in year one.
On cash alone, building overtakes the subscription at 3 seats.
Money you would actually spend
Time you would spend
—
What you would spend
What we assumed
The verdict above measures whether you could build it. This one is only about money.
Runnable build prompt
Build a minimal ChatGPT-like web assistant using Next.js (React) for the frontend, a Node/Express backend, Postgres for storage (or SQLite for single-container), and Docker. In scope: real-time chat UI with message input and streaming, backend endpoints that relay messages to an OpenAI-compatible API (accepts user-provided API key), persistent per-session chat history, simple session handling (guest + optional email login), retries and error handling for API calls, and automated tests for backend endpoints. Out of scope: reimplementing or training large proprietary models, plugin marketplace, image/voice generation, multi-tenant billing, or advanced moderation. Include logging, basic rate-limiting, health checks, and deployment Dockerfiles; provide setup scripts and a README with env var descriptions and a small test suite.
How we checked
How the score was reached
- Build verdict base78
- An open-source build was found+5
- 5 cited sources+3
- 3/3 assessment runs agreed+4
- Evidence score90
The base comes from the verdict. Everything under it is a check that either happened or did not, and each one is a fact frozen in this record rather than a judgement made at render time — so the same evidence always produces the same number.
How scoring works →Cited sources · 5
Every page the run actually retrieved.
- official productChatGPT product page
- official productChatGPT product page (features lines)
- official productChatGPT product page (privacy notice)
- open sourcearc53/DocsGPT
- open sourceaingdesk/AingDesk
Integrity checks
What held up, and what did not.
