ChatAndBuild launched Living Brain on 4 November, a memory layer that its founder Christel Buchanan describes as “one persistent brain for ChatGPT, Claude, Cursor, Codex and your own agents”. She posted on X: “We decoupled memory from the model”.
On its website, the company says that in internal tests Living Brain used 22.6 times fewer tokens per question than a full-context baseline. That figure comes from one benchmark, LongMemEval-S; on a second, LoCoMo, where it reports 92.2% accuracy, the site gives a 2.8 times reduction in context tokens.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on the model and the harness around it. It teaches the general idea only and does not discuss ChatAndBuild or Living Brain.
The key insight: ChatAndBuild’s Living Brain stores project context outside any one model so ChatGPT, Claude, Codex, Cursor and other agents can share it; its 22.6 times token saving comes from an internal LongMemEval-S test, while its LoCoMo run, the source of its 92.2% accuracy figure, shows a 2.8 times cut.
The site calls Living Brain “A persistent and self-learning interoperable memory layer”, available “via API, SDK, MCP”. It says Living Brain “stores knowledge outside the model and retrieves relevant context for each task”.
Wall St Engine summarised the use case on X: work through a project’s requirements in ChatGPT, save the decisions to Living Brain, then switch to Codex, which “retrieves the saved context, so you don’t have to copy everything across”.
Average tokens per question on LongMemEval-S, as ChatAndBuild states them from internal evaluations: about 5.1K for Living Brain against 116K for full context, the basis of its 22.6x claim. On LoCoMo, the site states a 2.8x reduction. Data: livingbrain.com, company claims, read 4 November 2026.
What Was Posted on X
ChatAndBuild is launching Living Brain.
Now you can start work in ChatGPT and continue in Claude or Codex without explaining your project all over again.
Itโs a shared memory layer that connected AI tools can read from and update.
For example: โ Work through a projectโs requirements in ChatGPT. โ Save the decisions to Living Brain. โ Switch to Codex to build it. โ Codex retrieves the saved context, so you donโt have to copy everything across.
Living Brain organizes the information and retrieves whatโs relevant to each task.
In internal tests, the company reports up to 22.6x fewer tokens per question versus its full-context baseline.
You can also add project updates through a ChatChat agent on WhatsApp, making them available to your connected tools.
Wall St Engine described the launch on X: start work in ChatGPT, save decisions to Living Brain, then continue in Codex, which retrieves the saved context. It relays the company’s internal claim of up to 22.6x fewer tokens per question.
ChatAndBuild founder Christel Buchanan announced Living Brain as one persistent brain for ChatGPT, Claude, Cursor, Codex and other agents, with the company’s own benchmark figures.
NEWS: Chat+Build is launching โLiving Brainโ which decouples memory from AI models.
One persistent brain that works across ChatGPT, Claude, Cursor, Codex, and custom agents. It allows users to write in any model and retrieve in any other.
Models become interchangeable compute; the memory stays with the user.
This is a big step toward agents that don’t reset every time you switch tools.
Exec Sum (@exec_sum) posted at 15:55 UTC on 4 November 2026 that Chat+Build is launching Living Brain, one persistent memory that works across ChatGPT, Claude, Cursor, Codex and custom agents, so users can write in one model and retrieve in another. It quotes Christel Buchanan’s (@ladyxtel) launch post, which lists 92.2% on LoCoMo, 464ms median retrieval and up to 22.6x fewer tokens per question; those figures are hers, not independently verified.
Two Benchmarks, Two Ratios
On LongMemEval-S, the site gives average tokens per question of about 5.1K for Living Brain against 116K for full context, and says the result was “95% lower cost per task versus a 116K-token full-context baseline”. By our arithmetic, 116K divided by 5.1K is about 22.7, in line with the stated 22.6 times.
On LoCoMo, the site says the latest internal run achieved “92.2% judged accuracy across 1,540 questions”, using “approximately 9.4K context tokens per question, compared with approximately 26K for the full-context baseline at 2.8ร fewer context tokens”. It reports a 464 ms median memory retrieval time on that run.
The site’s own note says: “Results are from internal evaluations”, and that “Savings depend on the workload and configuration”.
What It Costs
Living Brain’s pricing page lists an API plan at $8 per million input tokens and $0.50 per million output tokens, prepaid from $5, and a Workspace plan at $99 a month with three team seats. A Business tier is priced on request.
Why Portability Matters
Living Brain’s pitch is one memory that several assistants read from and write to, so a user can change tools mid-project. The site says it is compatible with “60+ AI Platforms”.
The Structural Read
As we read the launch, the product is portability: memory that belongs to the user’s project rather than to one assistant, reached through API, SDK and MCP.
As we read the numbers, the two benchmarks measure different things, and the larger saving sits on the test with a 116K-token full-context baseline; the company says savings depend on the workload and configuration.
The framework says: “Pricing power leaves the model and lands on whoever controls the surface and its reusable workflows”.
As we read it, a memory layer shared across assistants is a bid to own part of that surface, the project context, rather than the model. The framework predates this launch and does not discuss ChatAndBuild.
All benchmark figures are ChatAndBuild’s own internal results; the site says rankings apply only to the systems and configurations tested, and we have seen no independent evaluation. The company has not given user numbers or funding figures in the material we read.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
ChatAndBuild’s Living Brain is a memory layer meant to let ChatGPT, Claude, Codex, Cursor and other agents share one store of project context. Its 22.6 times token saving comes from an internal LongMemEval-S test; on the LoCoMo test behind its 92.2% accuracy figure, the stated saving is 2.8 times.
A note on sourcing. Benchmark figures are ChatAndBuild’s internal results as stated on livingbrain.com, read on 4 November 2026; prices are from its pricing page. The founder’s quotes are from her post on X; the workflow example is from Wall St Engine’s post on X. The 22.7 ratio is our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
