Writing With Voice Assistants

For years I took notes in meetings like my life depended on it. Typing every word. Missing every moment. I'd walk out with five pages of notes and still forget the one thing that actually mattered. Here's what finally clicked: taking notes isn't the same as paying attention. And it was never supposed to be your job. Your job is to be present. Let the machine handle the memory. Here's the 2-minute Claude workflow that takes notes for you, even if you never write a single word: 1. Install Your AI Notetaker Get Granola. It's the simplest AI notetaker out there. It runs quietly in the background on Zoom, Teams, or Meet. No bot joining the call. It just listens while you talk. 2. Join The Call And Forget It Don't take a single note. Don't type. Don't scramble. Just be present. Look people in the eye. Actually listen. This one shift changes how every meeting feels. 3. Open Claude, Not The Browser When the meeting ends, open the Claude app on your desktop. The app, not the browser tab. Click the "+" in the chat box and find Connectors. 4. Connect Your Tools In One Place Toggle on Granola, Gmail, and Slack together. Then pick the most powerful model and turn on extended thinking. Now Claude can see your meeting, your inbox, and your team chat all at once. 5. Pull And Summarize The Meeting Paste this: "Pull my last meeting from Granola. Summarize it in 5 bullets. List every action item assigned to me, with deadlines. Flag the 2-5 blockers I might hit." In seconds, the whole meeting becomes a clear plan. 6. Cross-Check Every Channel Paste this next: "Cross-check this with my Gmail thread and Slack discussion on this deal." Now nothing slips through the cracks between your tools. 7. Draft The Follow-Ups, Then Send Paste this: "Write the follow-up emails based on everything you just read. Save them as drafts in Gmail, not in the chat." Then review, tighten anything that needs your voice, and hit send. Meeting to action items in two minutes. The best operators aren't the ones with the most notes. They're the ones fully present in the room, because they built a system to remember for them. Stop transcribing your meetings. Start showing up to them. __ Enjoy this? ♻️ Repost it to your network and follow Matt Gray for more. Want to learn how a 30-minute voice memo becomes 30 days of content, leads, and clients? Join my free live Workshop on June 23rd (5 days away) to steal my homework: https://lnkd.in/eM7kshnK

This little use of AI has saved me literal hours every single week 🤭 I use three things together on every client call. Otter.ai Asana And whatever Google doc on we are working through with the client Whenever I jump on a call, I usually have the strategy document open and Asana open as well. Inside Asana we have a meeting project, and every single call has its own task so everything sits in one place. But the real game changer has been combining that with Otter.ai 🙂↔️ My Otter joins every call to record and transcribe everything that’s said (and I talk a lot x) We always try to reserve our client calls for creative brainstorming and strategic work together. Which means conversations go down lots of different routes and we end up covering a lot. And I have always felt that if you are trying to take detailed notes the whole time, your brain is multitasking a little tooooo much. So now during meetings I just leave quick markers. For example, if a client says something interesting about a launch idea, I might write something like: Minute 10 new launch idea Or if something needs following up: Minute 24 S mentioned sending XYZ. Add follow up task. Then, after the meeting, I go into Otter.ai. Each transcript has its own little AI chat box, so I can ask things like: - Can you summarise everything discussed around minute 10? And it pulls out the entire section of the conversation 🙂↔️ I used to record client calls and physically watch sections back afterwards, especially when a client was walking through something very specific or storytelling heavy. It used to take hours lol Now it takes minutes (I think this is where the kids would dab or something x) And I know some of you might be thinking - Margaret this is very obvious, shhhhh x But I also know that a lot of people I get on calls with have never even heard of using a transcriber like this. So friends, this is one small way I’m using AI at work right now x How are you using it 👀 It's not taboo anymore. I'm claiming that to be true

🎙️ Talk fast, write slow? You’re not alone—and AI might just be the missing link in your workflow. Some of my best ideas hit on the move—while walking, commuting, or between meetings. I used to lose them in forgotten voice notes… until AI stepped in. Now, with the right tools, I can: ✅ Transcribe voice memos into clean text ✅ Highlight key takeaways and structure outlines ✅ Turn spontaneous thoughts into social posts, blogs, or even pitch decks ✅ Skip the blank page and start from spoken momentum It’s like having an on-demand editor who speaks your language and formats your brainwaves into value. Talking is frictionless. Writing? Not always. But AI is helping bridge the gap between ideas and output. Have you ever turned a voice note into content using AI? I’d love to hear how you’re using tech to capture your creativity on the go. #AIforCreativity #ContentCreation #Productivity

Last weekend, I crashed my bike and broke my dominant elbow. Typing one-handed, hunt-and-peck style wasn’t going to cut it — so I started testing modern speech-to-text tools. A lot has changed... Seeing the rise of "vibe coding" — people building entire programs just by talking to AI made me wonder: are we finally at the point where we can do real, complex work purely by voice? This injury gave me the push to find out. Here’s what I’ve found so far: 🗣️ ChatGPT + Dictation (desktop or mobile interface) • Click the mic icon, speak, and lightly edit if needed. • Surprisingly good at punctuation, flow, and formatting. • Great for refining and tightening text before pasting into emails, documents, or anywhere else. • (By contrast, the dynamic speech mode is more conversational and less suited for deep work — no visible text.) 🖥️ Windows Voice Access • Built into Windows 11. • Manual activation often required; voice triggers are unreliable. • Handles direct input into any app — Slack, email, browser — no extra steps needed. • Punctuation, new lines, and capitalization need improvement, but it’s convenient for quick, rough entries. 📱 Apple iOS Dictation • Fast, intuitive, and extremely accurate — surprisingly on par with GPT dictation. • Works great inside mobile apps like Slack, Notes, and Mail. • Limited to mobile, but fantastic for short-form productivity & communications. Overall: speech-to-text is getting seriously good — especially when paired with AI that can clean up rough inputs. We're closer than ever to voice-driven workflows moving beyond accessibility into mainstream productivity. Right now, I’m mainly using ChatGPT Dictation, Windows Voice Access, and Apple iOS Dictation — and planning to explore Whisper API and Google Voice Typing next. 👉 Have you used speech-to-text seriously in your workflow yet? 👉 Which tools (or hacks) have made it actually work for you? Would love your suggestions — planning to pull these learnings into a deeper review soon! P.S. You might have guessed — this post itself was drafted with the help of GPT. I spoke the original ideas, and we iterated in conversation to refine it. Worked pretty well.

When your brain feels like it has 27 browser tabs open, try this 10-minute reset. This is my fastest way to get from spinning to one clear next move. Use it when you’re stuck, overthinking a client project, or starting from a blank page. 🦜 Step 1: Dictate (5–10 minutes) - Open ChatGPT on your phone and tap the mic (or use Voice Memo app). - Speak in full sentences. It's OK to ramble. Name the client, deliverable, blockers, half-ideas, worries. - End your recording with this line so AI knows what to do next: “That’s everything. Please sort what I just said into: (1) main themes, (2) decisions I need to make, and (3) one sensible next action.” 🧩 Step 2: Sort (1 minute) - If you used the voice memo app, paste the transcript into ChatGPT (iPhone now automatically provides transcripts, which is awesome!). - If you used the ChatGPT app, it will start doing it's thing. - You’ll get a short list of themes, a few decisions, and a single step you can take today. 🧵 Step 3: Pick one thread (2 minutes) - Scan the decisions. Choose the one that would unblock the most work. - Turn the suggested next action into a 25–30 minute task and put it on today’s plan. - Or if you need to work with the idea some more, get ChatGPT to help you unpack it further, review options, or turn it into a workable action plan. ❓ Why this works: - Talking is faster than typing. - Sometimes I think best when I'm out on a walk, which enables me to unload all my thoughts and ideas more effectively. - The “sort” instruction turns noise into buckets. - Choosing a single thread prevents you from rebuilding the mental knot. 🤔 Troubleshooting: - If the output feels vague, ask: “Make the next action smaller and more concrete.” - If it misses context, add 2–3 constraints (deadline, audience, tone) and rerun. - If you’re scattered, do a second 3-minute add-on and re-sort. 🧠 Alternate Prompt: If you forget how to end the voice recording (e.g., you're out on a walk and just want to talk it through for now), just type this prompt into the same chat once you're done talking (or once you're back home): “Turn this brain dump into: (1) 3–5 themes, (2) decisions I need to make with brief pros/cons, and (3) one next action I can finish in 25–30 minutes. Keep it concise and specific. If something’s missing, list the questions I should answer to move forward.” Enjoy!

My biggest productivity hack of 2024 hasn't been using ChatGPT, but rather a little known voice-to-text app called superwhisper. Back in the late '90s, I remember trying to use Kurzweil's VoicePad and thought that it would make my school work way easier just by speaking what I wanted to type. But it was clunky and far from accurate. After countless hours of frustration and garbled text, I gave up. But nearly 30 years later, I've found another tool Superwhisper by solo developer Neil C. actually delivers. Here’s what makes Superwhisper a game-changer for me: 🖥️ Reduced Typing: I can now write long pieces of content just by speaking. It's like the voice button on the Gboard Google keyboard but automatically cleans up the text with high-accuracy punctuation and casing. 🎯 Incredible Accuracy: Unlike the native Mac voice-to-text, Superwhisper’s accuracy is outstanding. It lets you run local transcription models and integrates seamlessly with online models like Deepgram, OpenAI, and Anthropic. It’s a night-and-day difference from my Kurzweil days. 💡 LLM Integration: The app can integrate with LLMs, so I can have my transcribed content polished up in no time. 📧 Everyday Use & Long-Form Content: Whether it's writing quick emails or creating detailed blog posts, product memos, or investment memos, Superwhisper handles it all just by listening to me talk. 🔍 Hidden Gem: Despite its power, it’s still a hidden gem with only 53 ratings on the App Store. It deserves way more recognition. 🛠️ Custom Vocab Arrays: It supports custom vocab arrays, which has been incredibly useful for accurately transcribing non-Anglo-American names and specific industry jargon. It also has translation support for 100+ languages, which isn't that useful for me, but maybe for you. 🌐 Translation Capabilities: It even has built-in translation capabilities and works with any app where you can paste from the clipboard. I don't often post reviews or endorse specific software, but I feel that I really need to call this one out, especially because it's been built by a solo developer and been so useful for me.

While cleaning my garage this weekend, I told ChatGPT Voice what I wished I had time to accomplish - it then launched a Codex task that operated autonomously for 79 hours and 48 minutes. 😳 I’ve had a few truly mind-blowing AI moments: ChatGPT was, of course, first. Then Claude Cowork the week it was released. I think ChatGPT's new Voice feature is the next breakthrough. In the last few weeks, Screentime says I've used ChatGPT 25–30 hours/week ON MY PHONE. Virtually all of that time is spent operating Codex, and probably half of it is through the new Voice mode, which is the first Voice model that can review open tasks, spin up new tasks, create sub-agents, call tools, etc. Seriously, if you haven't setup Codex on your desktop + controlled it with Voice while on a hike, driving, or garage cleaning, I highly recommend it. Here are a few tips I've learned: 1️⃣ ALWAYS delegate. Voice should be the chief of staff, not the worker. It's too slow and dumb to accomplish much itself. But it's excellent at writing good prompts to sub-agents. Keep brainstorming with Voice while other agents research, browse, inspect files, and build. Have Voice regularly check their status & steer their work. 2️⃣ Deploy to Sites or setup Tailscale. Any meaningful work requires you, the accountable human, to understand deeply wtf these agents are doing. Ask agents to turn proposals, architecture, and complex concepts into visual HTML explainers you can review visually from your phone via built-in Sites or a private Tailscale network. When I need to evaluate in bulk, I ask for swipeable cards, ranking interfaces, or a timer to gamify. Simple interface changes have a big impact. 3️⃣ Bookend the day. Morning: brain-dump, audit priorities, plan, delegate. Evening: close loops, audit my calendar (did I do what I said I would?), queue follow-ups, and shape up tasks that can run overnight. 4️⃣ Ask for outcomes that feel WAY too big. You're probably not getting enough value from agents because you're thinking too small. Make sure the agent builds ways to deterministically verify its own work (reducing subjective judgement), clarify the plan sufficiently upfront, and then let it run for hours or days. Being unshackled from my desk feels so liberating. It's got a long way to go but I can see a world where my phone is my primary work device, not my desktop. If you have used this & have things you want to add to my 4 tips above, please share!! I'd love to hear.

You think at ~400 words per minute. You speak at ~100 words per minute. You type at ~40 words per minute. The constraint isn't your ideas. It's the bridge between thinking and output. Fridays I share tools I actually use. Not sponsored. Not ones I tried once. This week: voice. I've dictated 30,287 words through whisperflow. According to their stats, that puts me in the top 3% of users. I believe it — I use it constantly. On walks. In the car. Waiting in line. Anywhere my hands aren't free but my mind is running. WhisperFlow handles transcription better than native iOS, which still stumbles on names, proper nouns, and anything faster than a careful dictation pace. But what I actually love is the small thing: it understands intent. If I say "let's meet at 4pm — actually make that 3:30" it types let's meet at 3:30pm. Not a literal transcript. A clean output. That handles most of what I need. But sometimes I don't want clean verbatim. I want to speak in stream-of-consciousness and have it formatted differently — as a structured email, a memo, a list of action items. Verbatim transcription can't do that. That's where Superwhisper comes in. Superwhisper lets you set output modes. You speak, choose a format — email, memo, SMS, whatever — and it runs the audio through local or cloud models to produce formatted output, not a transcript. Voice in. Structured content out. Two tools. One covers fast, accurate capture. The other closes the gap between raw thinking and polished output. What voice tools are you actually using? I'm always looking for what I'm missing — drop yours in the comments.

Quick productivity workflow I’ve been using for a while, in case it’s helpful for others. I have about an hour of commuting each day, and one of my biggest “hacks” has been turning that time into real focused work — not just podcasts, but actual progress on reading, writing, and learning. I generally think about it in three modes: 1) Reading long articles or documents If there’s something I want to get through (an essay, doc, spec, etc.), I’ll drop it into an iPhone app called ListenAI, which reads text out loud at whatever speed/voice I want. That makes it easy to “read” substantial material while driving. 2) Exploring, researching, and learning I also heavily use ChatGPT’s conversational voice mode for deep dives. For example, I’ll pass it a link and have it walk me through a research paper section by section, and I can keep asking questions until I fully understand it. It’s especially useful for going down tangents on background concepts the paper assumes you already know, then coming back — basically peeling back the onion from the 30,000-foot view down to the details. I’ll also use this mode when I’m trying to understand the “standard industry solutions” to a problem — AI is great for laying out the landscape of approaches and helping me get up to speed quickly. 3) Writing and drafting For document writing, I use voice input mode (different from conversational mode). This is ideal for long brain dumps: I’ll explain the context upfront (LinkedIn post, product spec, plan, etc.), the tone I want, and then just talk through my thoughts. We can speak much faster than we can type, and AI is extremely good at cleaning up messy raw thinking into something coherent and structured. One limitation: ChatGPT cuts you off around 10 minutes, so you sometimes have to do it in chunks. After it generates a draft, I’ll often paste it back into ListenAI to hear it read aloud at 1.5–1.8x speed (OpenAI’s read-back is still a bit flaky). Then I’ll go back in and give feedback to refine it. To keep all of this organized, I use TickTick for task management, including a dedicated “driving” list of things I want teed up for commute time. That way, even on a Monday morning before I’m fully context-loaded, I can glance at the list and immediately know what I can make progress on. Anyway — sharing in case any of these workflows are useful for others.