In partnership with
How do you give AI instructions now?
WTF is live AI coordination?: An agent starts doing the task.You can ask what it is doing, interrupt a bad decision and change the direction before the work finishes.
When ChatGPT became mainstream, AI was a place we visited.
You opened the chatbox, pasted an email and asked for a reply. Then you copied that reply into Gmail.
You did the same with messages for WhatsApp, Telegram and documents.
The model was useful. We were still the bridge between the AI chat and the real work.
Voice first saved us from typing every word. Then it gave us an actual conversation.
Now ChatGPT realtime is out and an agent can work and coordinate while we talk to it.
The important part is what happens after the task starts. We can catch a wrong decision and change the work before the agent finishes it.
I am Alex, welcome to ShortCu8 by Innov8.
Lets Dive Deep 🐰
⭐Today's Shortcut
AI now has four useful interfaces:
type when the instruction must be exact
dictate when the context is long and messy
use voice conversation to think through an idea
use live voice to steer work already in motion
All four remain useful. Each removes a different kind of friction.
We stopped trying to write the perfect prompt
Typing made us edit our thoughts before AI could see them. We often removed details while trying to compress a problem into a clean prompt.
Dictation gave us another option.
With tools such as Wispr Flow, you can hold a key and speak into almost any text field. Claude Code also offers /voice dictation, which turns speech into text inside its prompt box.
You can give the full context:
The dashboard broke after yesterday's update. The mobile view is fine. Do not touch the database because another team owns it. Check the API response first, then compare the desktop layout with the previous version.
Dictation is an established workflow, and it remains useful. Its boundary is clear: speak, transcribe, submit, wait. The agent starts after your turn ends.
Then AI became someone we could think aloud with
You can begin with an unfinished thought, hear the response and interrupt when AI misunderstands you:
No, wait. I am not asking for another content idea. I am trying to understand why this idea feels weak.
Claude Voice Mode is one example. It supports spoken conversations, interruptions and switching between voice and text while keeping the same conversation context. It can also use web search and connected tools under the user's plan and permissions.
This is different from Claude Code dictation. Voice Mode talks back. Dictation prepares text for submission.
Talking also exposes uncertainty. You can change your mind halfway through a sentence and explore something before you know the right prompt.
The AI is still mainly responding to the current conversation. The work usually waits for the next turn.
Now the conversation can continue inside the work
An agent is rewriting a presentation. It has opened the files and started changing the slides, but it is writing for customers even though the deck is for investors.
With live coordination, you can say:
Stop changing the product slides. Keep the numbers. Rewrite the story for investors and show me the new outline before editing anything else.
ChatGPT Voice in Work and Codex is a shipped example of this direction. OpenAI says Voice can start tasks, check progress, interrupt or redirect work and coordinate agents using the tools and permissions available in Work or Codex.
While an agent researches, ask which original sources it found and reject weak ones.
While it edits a document, keep one section and redirect another.
While it codes, tell it to leave the database alone and fix the API.
You are no longer carrying every update into a separate chat. The correction reaches the agent while the work is still moving.
Voice in Work and Codex currently requires an eligible desktop account, and only one Voice conversation can run at a time.
We will still type
Talking out loud is awkward in a shared office. Code, URLs, private information and exact commands are better handled in text.
Typing gives work a durable record:
the approved requirement
the exact change requested
the final decision
the command that may affect production
Use dictation for a large amount of context, voice conversation for exploration and live voice for steering an active task. Put commitments, approvals and final instructions back into writing.
Why this needs a boundary
A natural voice can feel more trustworthy even when the underlying answer is wrong.
A study of 3,500 participants across ten countries found that conversation flow and response speed affected how humanlike a chatbot felt. Making the chatbot feel more human did not create a universal increase in trust or engagement.
So voice should make correction easier, not review optional.
Now go and try codex voice mode 🫨
The ShortList
🛠️Cool Tools of the Week:
Qwen-Image-3.0: New model from Alibaba, the third generation of its foundational image generation model.
Runway Media Router: The AI video startup launched a tool that automatically selects the best image, video, or audio generation model.
ChatGPT Health: OpenAI's health tool is now available to all US users over 18.
FLUX 3: The latest model from Black Forest Labs, capable of generating images, video and audio, as well as use in robotic action.
Gemini Spark: Google expands access to personal AI agent Gemini Spark.
📩 Innathe Shortcu8 engane undarunnu 👇️?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha Shortcu8il kanaam bie…👋
If you read till here, you might find this interesting
#AD1
Win AI Search Without a Big Team
92% of VCs use AI to find companies. 58% of buyers start there, too. If you're not showing up in AI answers, you're invisible before the conversation even starts. Join HubSpot for Startups, Anthropic, and Marketing Against the Grain on July 30 (11 am ET) for a live AEO teardown. Real startup. Real recs. Register and unlock the free Startup Visibility Bundle.
#AD2
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.