Memoket

Turns out, getting AI to write a good summary is harder than telling it to “summarize this.” When we tested different approaches internally, we got very different feedback. One person wanted something shorter. Another wanted more context. Someone else wanted the same information, just easier to read. And they all had a point. That's part of why we built different summary formats into Memoket: Brief, Summary, and One Pager. We also worked on how AI decides what to include, how to structure it, and which details actually matter. A little behind-the-scenes look at how we designed Memoket's summaries 👇

If you're connecting Memoket to another tool, choose the event for the thing you need. recording.uploaded means the audio has arrived. The transcript and summary aren't ready at that point. summary.completed is the event to use when your next step needs the summary. There is another detail worth handling before a workflow goes live: a webhook delivery can arrive more than once. Store the event ID and skip one you've already processed. Otherwise a retry can create a second document or follow-up draft. And a newly generated summary produces a new completion event. Decide whether your destination should update the existing item or create another version.

👀👀👀👀👀

Need to put a conversation brief into a team chat? Memoket can give you smart meeting followups, and export a One-Pager as an image. Before sharing, open the image at phone size. If the recipient has to zoom in to read the decision, use a shorter brief or send a text/document version instead.

Revisit conversations, explore context, and find your next steps, all in one view with Memoket on iPhone Duo.📲📱

Before buying any AI wearable, ask: can you go back to the original recording? Some products prioritize processing conversations into notes without retaining the audio. Others let you keep and replay the recording. Neither approach is automatically right for everyone. If you're using notes to recall the general discussion, a summary might be enough. If you need to check an exact quote, an unfamiliar name, or something the transcription misunderstood, access to the audio becomes much more important. We explored this and other differences in our Bee Frontier vs. Memoket Gem comparison. memoket.ai/blogs/news/memoke…

In Microsoft's Bankwest case study, the bank's AI assistant handles simple customer questions and passes complex ones to staff with the conversation history. One small detail stood out: staff suggested push notifications that take customers directly back to their existing chat. You can build a good AI assistant and still make people hunt for the answer when they return. For anyone building AI into a product, test what happens after the user leaves and comes back. Does the notification open the conversation they need, or send them somewhere they have to start searching? The experience doesn't end when the AI finishes its job!

A respondent in our user survey records daily sprint stand-ups, exports the transcript to Claude and uses it to draft follow-up emails. Here is a prompt worth trying after that export: “Draft the follow-up for someone who missed this stand-up. Separate requests for help from progress updates. Leave out anything that doesn't affect their work.” Read the draft against the transcript before sending it. A blocker mentioned in passing can otherwise become an instruction addressed to the wrong person.

A voice note after a customer call can quietly turn your interpretation into something the customer supposedly said. “I think price is the problem” becomes “the customer objected to the price.” Those leave a teammate with different jobs: test a possibility, or solve a stated objection. In a spoken debrief, leave the distinction audible: “They asked about payment terms. My guess is that cash flow is the concern. I haven't checked that with them.” Then ask for the follow-up note to keep those parts separate. Include the question you still need to ask. Someone picking up the account should be able to tell where the customer's words end and your reading of them begins.

Apple Calendar 🤝 Memoket You can now connect your Apple Calendar to Memoket! 📅 Another way to bring your schedule and conversations together in one place. The integration is now available, so give it a try and let us know what you think! #Memoket #AppleCalendar

In our October survey, 5 of 28 respondents had used MemoChat. If you've opened it but weren't sure what to ask, what was on the screen that left you stuck? If you do use it, what was your first useful question? 👀

Choosing an AI recorder isn't just about transcription accuracy. Plaud NotePin and Memoket Gem both help capture conversations, but the hardware, recording workflow, and what happens to your notes afterward are different. If you're comparing wearable AI recorders, we'd start with three questions: • How will you actually wear and use it? • What can you do with the recordings afterward? • What does the ongoing subscription cost look like? We compared the two across hardware, AI features, and pricing.

Some of the work inside an AI memory system is a very small question asked a lot of times: is this passage useful enough to keep searching with? Hindsight's new decision model handles judgments like relevance and whether there is enough evidence. The LLM still writes the answer. That division is worth paying attention to. A choice between a few options doesn't always need a paragraph of generated reasoning, followed by code trying to parse the paragraph back into a choice. The tradeoff is in the gate: reject useful evidence early and the writer never gets to see it. We'd want to inspect those rejected passages alongside the cost savings.

This video covers trip planning - but one Gem user reads out cookbook ingredients, adjusts the quantities for the people eating, then adds what is already in the fridge. MemoChat turns that into the shopping list. That last bit is important because a recipe knows what goes into dinner. It doesn't know what you already bought. Try a small version: record one recipe and the ingredients you have. Ask MemoChat for the remaining shopping list, then check the quantities before heading out. Let us know how it goes!

Your AI note taker captures every Zoom meeting. But what about the conversations that happen after the call ends? A quick discussion with a colleague. A client meeting over coffee. A decision made in the hallway. That's one of the biggest differences between Memoket and Granola. Granola focuses on AI-powered meeting notes. Memoket combines virtual meeting capture with a wearable for in-person conversations, bringing everything into one searchable conversation history. We put together a comparison covering features, supported platforms, pricing, and where each tool makes more sense. memoket.ai/blogs/news/memoke…

“The client conversation” is a weak cue when you've had twelve of them. “Restoring the upstairs flat before the tenant moves back” gives you more to work with: a place, a job and a reason the person called. Our Memory Anatomy article has an interactive ACT-R model showing how use and context can affect what comes to mind. One useful takeaway for your own notes: preserve a distinctive phrase you could imagine searching for later. It doesn't have to become an elaborate filing system. “Upstairs flat, water damage, tenant moving back” is already a better starting point than “call notes.” memoket.ai/blogs/news/memory…

Already doing the next piece of work in ChatGPT? Memoket Desktop has a ChatGPT connection for using your conversation context there. A practical demo: bring in a harmless project discussion, then ask ChatGPT to draft a short explanation for someone who missed it. Keep the original conversation beside the draft while you check it. No client names or confidential material needed to try the workflow.

If the office coffee machine could transcribe our conversations, half the action items would be "we should really get back to work." ☕️

A founder on Reddit noticed a habit after asking AI to examine their conversation transcripts: they explained the whole chain of reasoning before getting to the point. Sometimes the other person only needed the answer. They were using Claude or ChatGPT manually, not a dedicated coaching feature. And they were already wary of AI being too agreeable. One useful suggestion in the discussion: hide the dates on two comparable conversations. Label them A and B. Ask which gets to the point sooner, and ask for the passages that support the answer. Then reveal which was newer. That makes “you've improved” a little harder to produce just because one file says October and the other says September.

Friday afternoon. You've had a dozen meetings, several project updates, and a few decisions that changed along the way. Now you need to put together your weekly report. In MemoChat, open Report and bring together the relevant conversations from the week. Memoket can organize that context into a structured report covering progress, decisions, action items, and outstanding questions. This demo uses fictional conversations to show how it works.