Apple Might Finally Build the Voice Assistant We Deserve

The funny thing is, most tasks that the average person performs can already be automated one way or another, and all of this can technically be implemented for specific use cases.

Overall this resembles what a voice assistant was supposed to be. All these Alexas, Siris — they were supposed to be exactly like this, and now we’re reaching the point where it’s actually possible.

And it feels like Apple could release this very soon. Because they have the whole ecosystem tied to an Apple account, which includes photos, emails, messengers, the operating system, and everything else.

Why We Need a Voice OS: Beyond the Cursor Interface

For the past couple of weeks I’ve been actively working on a speech recognition and dictation app. Along the way I keep getting ideas for new features. For example, I immediately added an output processing option where you can enter a prompt and it’ll run your transcript through GPT.

Spinning up all these ideas in my head, seasoning them with content from my info feed, I came to the conclusion that the voice OS concept — or a voice operating system — isn’t actually crazy!

I’ve noticed that many people are starting to use Cursor as the central hub for organizing their life and business. They store information about themselves, their projects, connect various MCPs and agents, fully managing their lives through this interface. Example 1 and example two.

But Cursor’s current interface is unnecessary. It shouldn’t be like this. Users don’t need all that scaffolding. You shouldn’t have to launch a separate project in Cursor to get started — it rips you out of context.

When I started dictating more with my voice, I got the urge to press fewer buttons, got lazy about opening files and programs. From there comes the idea of a voice assistant built into the system. If you develop that thought, it becomes clear that we don’t really need the whole interface, except for situations that require visual interaction.

As a user, I want to always have access to my database and knowledge. A simple dictation and speech recognition app can easily become a tool where you can give commands and connect MCPs and agents. You can say “in such-and-such project add such-and-such task” and overall the agent has enough intelligence and tools to find your project and make some changes there.

I want to automate managing various interfaces and agents and make it accessible through natural human interfaces. We can’t create universal physical buttons to control everything with our hands, but we can already control many tasks with our voice.

For example, I’d like to reply to emails by voice. Sure, I open the email client with a mouse, but then I don’t need to click through folders, search for emails. It’s enough to tell the voice agent to open an email from a specific person and draft a reply, schedule a meeting — that would be perfect.

The Voice OS operating system concept fits this vision of the future. Curious what you think about this, cool or sketchy?

Programmers Then vs Now: The Real Reason We Can't Sleep

Programmers back in the day: Can’t sleep because they’re thinking about what algorithm and logic they need to implement

Programmers now: Can’t sleep because they’re thinking about what prompt they need to write

Why Google Antigravity's Voice Input Changed My Vibe Coding Game

Voice Input in Antigravity

For the last six months I’ve been deep into vibe coding: about once a month I sit down and spend several days grinding on some genius idea. Nothing solid has come out of it yet. Right now I’m working on a voice recognition tool so I can easily dictate messages whenever I want.

I realized I don’t like Cursor. I used it for quite a while, I had a paid subscription. Then I switched to OpenAI Codex and played around with it for a long time. I installed the plugin and canceled my Cursor subscription because I bought an OpenAI one…

And now that Google Antigravity is out, it’s an absolute beast. The default limits are super generous. Gemini Flash works great, and it’s the cheapest and fastest model. Plus, it has built-in voice recognition, so you can dictate tasks to the agent right in the editor.

This is a straight-up killer feature because for some reason Cursor doesn’t have it. Without a voice recorder it becomes a huge barrier. I used to constantly stress about what to write to the agent… I tried writing in English, switched to Russian, and now in Antigravity I just hit the voice button.

I just dump all the chaos from my head — and that’s enough: it understands everything perfectly. Nothing else needed. I can do other things on the fly, drink tea, play guitar at the same time. That’s the huge value of Antigravity — highly recommend trying it.

Right now I’m making a voice recognition app — analogous to WhisperFlow or MacWhisper. There are Mac apps that can transcribe voice to text. And now I’m making a free alternative, want to open-source it.

I don’t know if something like this already exists — I didn’t find anything in a quick search. When I was looking for a voice-to-file transcription tool, I got so fed up that I had to vibe-code my own on Lovable. Anyway, I’m working on it. Will share soon.

This text I dictated into my Swift app 💪

I Can't Stay Silent: AI Slop Is Meme Culture 3.0

I can’t take it anymore! You need to know this:

I love AI slop and everything the latest video models are putting out. This is literally memes 3.0!!11 An endless source of laughs! A breath of fresh air!

I’m watching all these videos with great interest.

Drop your favorite clips, please