Browser-native · select → understand

What does it
mean to me?

WDIMTM is an AI explainer for the moment you hit something confusing on the web. Stay on the page. Select the text. Get a short answer that respects your context — not a generic chatbot dump.

Install the extension View source

Chrome MV3 extension BYOK OpenAI-compatible or Claude No prompt required

docs.example.com/caching

Hot keys & distributed caches

Memcached is another distributed, in-memory key–value cache, often used as a simpler alternative to Redis.

Switching from Redis to Memcached does not automatically solve a hot-key problem: a single extremely popular key can overload the one cache node responsible for it.

Typical mitigations include replicating the value across nodes, splitting one logical key into multiple keys, client-side caching, or request coalescing.

? Explain Engineering ▾

The problem

Chatbots make you leave the moment

Most AI tools force a detour: copy, switch app, invent a prompt, paste, wait. WDIMTM is built for the opposite — stay in the flow of reading.

Typical AI workflow

1 See something confusing
2 Copy text
3 Switch to ChatGPT / Claude
4 Write a prompt
5 Paste context again…

WDIMTM

1 See something confusing
2 Select it on the page
3 Understand — with your lens & memory

How it works

One selection. Your lens. Personal relevance.

Page facts, your profile, relevant memories and the active lens compose into one OpenAI-compatible request — returning a concise explanation, predicted follow-ups, and an optional memory suggestion. When a short answer is not enough, the same context escalates into a conversation.

Bounded page context

Uses the selection and nearby text — never the full DOM. Fast, private, and grounded in what you’re actually reading.

Lenses, not prompts

Explain, Sanity check, Fact check, Opportunities, Engineering, Investing. WDIMTM can suggest the lens from the selection, or you pin one and it stops guessing.

What it means to you

Optional profile + memories shape implications. No forced personalization — if nothing personal applies, you still get a clean explanation.

From answer to conversation

Follow-up chips are derived from the content, not the same three buttons forever. When one tap isn’t enough, Discuss further opens a page chat carrying the same selection and context.

Live evidence, only when asked

Verify, Research and page chat can pull fresh web snippets through your own Tavily, Brave or Serper key. Ordinary Explain stays offline and single-shot.

User-controlled memory

Suggestions only. Nothing is saved until you confirm. Inspect, edit, forget, or switch memory off entirely.

Start from what you already said

Import a ChatGPT or Claude export and WDIMTM distills it into candidate memories. Reviewed one by one before anything is saved, read in the page only, never written to disk, never sent to a server of ours.

Screenshots are questions too

Take a system screenshot and paste it straight into the page chat — or upload, or drag it in. No text prompt required, and no screen-capture permission: the extension never records your screen itself.

Off where you want it off

One switch in the toolbar turns WDIMTM off for the site you are on, and the denylist keeps it off. Nothing is read on a site you excluded — the content script never activates there.

Lenses

Same text. Different questions.

A fee switch headline means one thing to an engineer and another to an investor. Lenses encode that without making you prompt-engineer.

Explain Plain-language unpacking Sanity check Logic, plausibility, missing pieces Fact check Checkable claims and what would verify them Opportunities Incentives, mispricing, monetisation Engineering Architecture, trade-offs, failure modes Investing Risk, thesis, competitive dynamics

…plus custom lenses written in your own words.

Access

You bring the model. We bring the moment.

WDIMTM owns context and personalization — not your GPU bill. Choose how inference is paid for.

Recommended

Bring your own key

Any OpenAI-compatible API — OpenAI, OpenRouter, a local model, your own gateway — or Anthropic directly with a Claude key. One request per explain, streaming supported, and you pay the provider rather than us.

Runtime ids: openai-compatible · anthropic

Not open yet

WDIMTM Cloud

An optional hosted mode for people who would rather not hold a key: managed inference, memory that syncs across devices, and research that keeps running after the tab closes. Same extension, same answers — it changes who pays and where the key lives. Local and bring-your-own-key stay free, permanently.

Runtime id: wdimtm-cloud

Install

Build it and load it

Not on the Chrome Web Store yet. Install from source — Chrome or any Chromium, developer mode, about three minutes.

Clone and build

git clone https://github.com/wdimtm/wdimtm-extension.git

Then npm install && npm run build. The loadable extension is dist/, which the build produces — there is nothing to load before you run it.

Load the extension

Open chrome://extensions → enable Developer mode → Load unpacked → select the dist/ folder.

Add your key

Open WDIMTM Options → AI access → Use my own API key, and pick an OpenAI-compatible endpoint or Anthropic. Hit Test connection, save, then select text on any page. Until you do this you get sample answers, clearly labelled as such.