Skip to the content
Product Weaver

Changelog

What changed in each release of the Product Weaver plugin, newest first. The full notes are in the plugin's CHANGELOG on GitHub. To update, run claude plugin update productweaver@productweaver-plugin in Claude Code, or update the marketplace in your assistant.

0.12.0

  • Your map, review sheet and interview guide now go to a pw-output/ folder you can see in Finder, next to the hidden .pw-kb/ study folder. The next run moves them there by itself.
  • Each run ends with one link, or a file card where your app shows one, that opens the map. The interview guide is in its Next interviews tab, ready to print or copy.

0.11.4

  • In Cowork, a short drop of the link to your computer no longer stops the study: the assistant waits and tries again by itself, and if the link stays down, it goes on from where it stopped when you say Try again.

0.11.3

  • In Cowork, the study is saved in your folder: with a folder connected to the task or its project, the map, the review sheet and the study go into its .pw-kb folder instead of the chat, and a later task finds the study there.

0.11.2

  • Claude Desktop is named correctly in usage statistics: a Cowork task, the Code tab and a chat each use their own name.
  • The usage README says how to let Claude Desktop send: allow www.productweaver.com under Settings, Capabilities. It also now says it reads git's index to tell whether git tracks the run log.

0.11.1

  • No more crash where Python has an old jsonschema package (before 4.0): the plugin now uses its own built-in check there.

0.11.0

  • Usage statistics, only if you say yes: the assistant asks each person once per study. Without a yes, nothing is recorded or sent.
  • After a yes, small events (which steps ran, counts, error codes, versions) go to productweaver.com, never your interviews, names, quotes, prompts or file names.
  • See every event in .pw-kb/usage/sent.jsonl. Stop by saying “stop sharing”, or set KB_TOOL_USAGE=0 or DO_NOT_TRACK=1.

0.10.0

  • A new skill, pw-start: say “Help me get started with Product Weaver” and it turns your transcripts into a discovery map, or, with none yet, writes a guide for your first round of interviews.
  • Back in a study, ask “What should I do next?” and get one step, with the words to say. Every report now ends with one next step.
  • In chats that forget their files, such as ChatGPT, the assistant hands back the whole study to upload with your next interviews.

0.9.0

  • A Next interviews view in the discovery map plans your next round: who to talk to, what to ask, and how many more conversations make you surer.
  • The assistant also writes interview-guide.md, a guide to take into the calls.
  • The Progress tab starts with how sure you can be, and its chart no longer covers the text under it.

0.8.0

  • Runs with Python alone: nothing to download, so it also works where the assistant's sandbox has no internet access, such as ChatGPT Work.
  • The study's files are now JSON. A study made with 0.7.0 or earlier is converted once, the first time you use it.
  • Each discovery map is about 110 KB smaller.

0.7.0

  • A study can be in the interviews' language: when the interviews are not in English, the assistant asks once whether to write the study in their language or in English.
  • Everything you read follows that choice: the findings, the review sheet and the map's own buttons and help. Quotes stay in the words people used.

0.6.0

  • The Overview starts with who the study is about: an icon, a short name, one sentence, and the people you talked to.
  • Plainer words in the Overview, and the assistant's own sentences are checked for jargon.
  • The map says which plugin version made it, and when.

0.5.1

  • Calmer navigation in the discovery map: scroll moves the map, and pinch (or Ctrl or ⌘ with the wheel) zooms.
  • Each kind of box on the map has its own look: a step card, a green measure of success, and a pill for each solution.

0.5.0

  • The skill is now called pw-discovery (it was distill-kb). Asking in plain words works as before, and existing studies need no change.

0.4.0

  • The discovery map opens on an Overview: a short story in six steps, from who you listened to, to where a new solution should focus.
  • A “What to do next” block with three tracks: keep learning, prototype, and go to market.
  • Value themes group a core job's desired outcomes into a few needs in the customer's voice, with a Value themes view in the map.

0.3.1

  • The first discovery map of a study opens in your browser by itself.

0.3.0

  • Everything a study needs lives in one .pw-kb/ folder, with a copy of each transcript the knowledge base quotes.
  • Current solutions: what people use today at each step, grouped by the work they compete for, including what they stopped using and why.
  • A Solutions view in the discovery map, and “Ask next” names at most five thin steps per core job.

0.2.0

  • A study can hold more than one core job, each with its own job map.
  • Every interview is linked to the core job it rolls up to.

0.1.2

  • Exact package versions and a lock file, so every machine installs the same packages.

0.1.1

  • Changes for the Anthropic directory review.

0.1.0

  • First release: customer interview transcripts become an Outcome-Driven Innovation knowledge base in one YAML file, with every quote checked against its transcript.
  • A review sheet and an interactive discovery map that works offline.