The Making of Claude Code
Anthropic’s oral history traces Claude Code from early coding assistants and the internal clide CLI to a public product. Lessons in model capability, product design and feedback.
Anthropic’s oral history of Claude Code is useful because it describes several experiments rather than one inevitable product plan. Early coding assistants, internal tools, model research and a later CLI product contributed different pieces.
The account is told by researchers, engineers, product designers and early users. Their recollections are primary testimony about the project, not an independent measurement of every adoption or productivity claim.
Coding research before the product
Anthropic’s contributors describe work on coding models from 2021 and an early VS Code coding assistant in 2022. They also describe uncertainty about turning research into a product. That is narrower than saying the company had a formal rule against selling products.
The interesting tension is between research capability and a usable tool. A model can show promise before its surrounding environment is reliable enough for everyday development.
clide: the tool before the tool
The oral history distinguishes two tools. The early coding assistant was a VS Code extension; clide was an internal command-line tool for code editing and development tasks. Treating those as the same interface obscures the sequence.
This is the part worth sitting with. The tool earned its right to exist internally, through use, long before anyone tried to turn it into a product. Nobody had to be convinced by a deck. The value showed up in the daily work of the people building the models.
Boris Cherny and the two-week sprint
Ben Mann dates Boris Cherny’s arrival on Labs to September 2024. Sid Bidasaria describes a December sprint that added core product features, while Meaghan Choi recalls a separate two-week design effort. These recollections describe focused work on top of earlier experimentation.
Early reception and public launch
Cat Wu describes lukewarm feedback during the early-access program before the February 2025 public launch. That is not the same as evidence that the public launch itself failed.
Early feedback can reveal a real idea inside an unreliable product. The practical lesson is to distinguish user demand, model capability and execution quality instead of treating one launch reaction as a verdict on all three.
Capability and packaging changed together
Meaghan Choi identifies the Claude 4 models as an important capability threshold. Boris Cherny also credits subscriptions. The account therefore supports a combination of model improvement and product economics, rather than a model-only explanation.
For builders, that suggests tracking both technical acceptance and adoption conditions. A better model can expand what an interface can do, while access, price and usability determine whether people can use it repeatedly.
The timeline in one view
| Period | Account in the oral history |
|---|---|
| 2021–2022 | Coding research and an early VS Code assistant |
| Before Claude Code | Internal command-line tool clide |
| September–December 2024 | Labs work, Claude CLI iteration and product/design sprints |
| February 2025 | Public Claude Code launch after early-access feedback |
| Later in 2025 | Contributors credit Claude 4 and subscriptions for greater adoption |
What builders should actually take from this
The nostalgia is nice. The lessons are more useful.
The uncomfortable takeaway
The durable lesson is iterative fit between a model, an execution environment and a user workflow. Internal use supplies feedback, but external users and independent acceptance checks remain necessary.
An oral history explains how its participants understood the process. Use it to form design hypotheses, then test those hypotheses in the environment you are building.
Sources: The Making of Claude Code — Anthropic oral history. Documentation checked September 20, 2026.
