If you’ve been following the AI space lately, you’ve probably noticed something shifting. We’re moving past the era of chatbots that just answer questions. What we’re seeing now — especially in today’s batch of launches — are AI tools that actually do things. They design interfaces, optimize delivery routes, interview candidates, and even sit inside Slack acting like a real team member.
I went through all of today’s releases and picked the ten that genuinely stood out. Here they are.
Wispr Flow
Wispr Flow been in beta for months and people wouldn’t shut up about it. Now it’s out. You press a key on your keyboard, talk like you’re talking to a human, and it types wherever your cursor is. Slack, email, Figma, terminal, doesn’t matter. It learns how you write over time so it doesn’t sound like a robot took dictation. And if you’re at a coffee shop or sitting next to someone, there’s a whisper mode that picks up quiet speech without bothering anybody. Been testing it for a week and I already feel slower when I have to type.
Key Features:
- Responds instantaneously with a single keystroke, eliminating delays
- Learns your writing style and automatically adapts its tone accordingly
- Whisper mode enables usage in quiet environments without disruption
Stitch 2.0 by Google
Stitch 2.0 is answer to the whole “vibe coding” thing that’s been everywhere. You type what you want—like “dark mode dashboard for fitness tracking with rounded cards and purple accents”—and it just builds it. Not a mockup. Not a wireframe. A real, production-ready interface with code you could actually ship. The designers I showed this to had mixed feelings. The developers were all smiles.
Key Features:
- Generates complete user interfaces within seconds from simple descriptions
- Produces clean, deployment-ready code
- Integrates seamlessly with existing design tools and workflows
MiniMax-M2.7
MiniMax-M2.7 is for the nerds. It’s an AI model built specifically for agents that need to run on their own, not just answer chat questions. And it gets better the more you use it without you having to retrain anything. It’s open source too, so you can dig into it and change whatever you want. The architecture is different from the usual stuff out there—it was designed for autonomy from day one.
Key Features:
- Improves autonomously through usage without requiring retraining
- The open-source codebase allows developers to customize freely.
- Architecture optimized explicitly for agents rather than conventional chat applications
Krisp Accent Conversion
Krisp already made noise cancellation feel like magic. Now they’re tackling accents. I jumped on a call with someone from Scotland last week and it just worked. No delay, no weird artifacts. The other person sounded clear the whole time. If you work with teams spread across different countries, this is one of those things you don’t know you need until you use it and then you can’t go back.
Key Features:
- Eliminates communication barriers within global teams
- Operates with imperceptible latency during live conversations
- Enhances inclusivity for non-native speakers in professional settings
InfrOS
Every developer has a story about deploying something to the cloud and getting a bill that made their eyes water. InfrOS is supposed to fix that. You run your architecture through it before you write any code and it tells you what’s going to cost what, where things might break, and whether your security setup actually makes sense. It’s like having a senior DevOps person look over your shoulder before you commit to anything stupid.
Key Features:
- Accurately estimates costs and performance during the planning phase
- Identifies bottlenecks and security vulnerabilities prior to launch
- Integrates with DevOps pipelines for continuous validation
Netlify.new
Netlify has been around forever. But this new thing they released is actually kind of wild. You type what you want to build, hit enter, and it spins up a full project with a live URL in under a minute. I tried it with a simple portfolio idea and had something running before I finished my coffee. It’s perfect for those 2am ideas where you just want to see if something works before committing to building it properly.
Key Features:
- Creates and deploys projects from straightforward descriptions
- Reduces time from conception to publication to minutes
- Leverages Netlify’s entire infrastructure automatically
Unblocked
Unblocked gives your coding agents the context they’re usually missing—the history, decisions, and conventions behind your codebase. Works with Claude Code, Cursor, VS Code, any agent workflow. A staff engineer I know called it “the holy grail” because his agents finally stopped hallucinating and started pulling from everything his team had ever worked on.
Key Features:
- Connects code, docs, and conversations to give agents real-time organizational context
- Cuts token usage by nearly 50% and speeds up tasks by over 80%
- Works with any agent workflow via MCP, CLI, or API
getviktor.com
Getviktor it’s an AI that lives inside your Slack and just watches how your team works. After a while it starts offering to automate things you didn’t even realize you were doing manually. It’ll notice patterns and suggest stuff. The weird thing is it actually works. A friend who runs a small agency said it’s like having an intern who’s always paying attention and never needs to be told twice.
Key Features:
- Suggests automations users hadn’t previously considered necessary
- Executes complete tasks without requiring explicit commands
- Learns from team interactions and adapts organically to workflows
OpenAdapter
OpenAdapter is one subscription that gives you every open-source model worth using—DeepSeek, Qwen, Mistral, Kimi K2.5, MiniMax-M2.7, all of them. Works in Cursor, VS Code, Windsurf, any tool that speaks OpenAI’s API. It’s about 4x cheaper than Claude Code. A friend who codes for a living switched last month and said he spent more time managing keys than actually building things.
Key Features:
- One subscription gives access to 74+ open-source SOTA models
- Works with any OpenAI-compatible editor or tool—no extensions needed
- Predictable pricing starting at $9/mo, no surprise API bills
Framer
Framer’s been around for years but the new version finally feels like what they always promised. You build a site, launch it, and it just scales. No worrying about traffic spikes. No upgrading plans. The AI helps you design and publish at the same time, which sounds chaotic but somehow works. A friend used it to launch her consulting site and had it live in an afternoon. Looked like something an agency would charge five figures for.
Key Features?
- Integrated AI assists in simultaneous creation and publication
- Scales automatically according to business growth
- Ideal for teams requiring agility without compromising quality
A Quick Takeaway
Look, I’ve been covering AI launches for a while now, and what struck me about today’s batch is how practical everything feels. Nobody’s promising artificial general intelligence or robots taking over the world. Instead, we’re getting tools that solve real, annoying problems: miscommunication due to accents, slow cloud deployments, endless design iterations, and recruitment processes that take weeks.
The throughline here is autonomy. And that shift, from copilot to coworker, might be the most interesting development we’ve seen all year.
Read Also
I Launches: Top 10 of May 22, 2025
Take a look back at the releases that defined last year, including Macaly (app creation through conversation), Cua (open source agents), and Mitsuko (AI-powered subtitle translation). A helpful retrospective to understand how the market has evolved.
