I think I’m finally ready to let other people into what I’ve been building. This is rootmail.io I’ve spent a long time building it, breaking things, rebuilding them, changing my mind, and discovering that some ideas I was certain would work… didn’t. But this is the first version I feel comfortable putting in front of other people and saying: Come in. Try it. Tell me what happens. Rootmail is a place to send email and understand what happens after you hit send. You can write and send directly from the dashboard, or connect it to something you’re building through code. Your templates, audiences, delivery history, replies, and sending activity stay connected in one place. And if you’re building a product that sends email on behalf of different customers, Rootmail is being designed to let you manage those senders separately without turning your email infrastructure into a mess. But I’ve reached a point where using it myself can only teach me so much. Now I want to see what you do with it. Where do you get stuck? What feels surprisingly useful? What feels confusing? What makes you stop and think: “Why can’t it just do this?” That is the part I care about most right now. There is one important limitation I want to be transparent about. Rootmail is currently operating within AWS’s restricted sending environment. During this beta, you’ll need to verify the email addresses you want to test with, and sending volumes will remain intentionally small. So this is not the moment to move a 50,000-person mailing list over. It is the moment to bring one real thing you want to send. Maybe: An email for a product you’re building. A transactional notification. A template you keep recreating. A small customer workflow. A reply flow you want to test. Or simply an email process that currently feels more complicated than it should. Send it through Rootmail. Reply to it. Follow what happens. And tell me how it went — even if the feedback is simply: “I got lost here.” That feedback is incredibly valuable. The beta also helps us build a record of responsible, real-world usage as we work toward broader sending access with AWS. AWS ultimately controls that approval, but genuine usage gives us something concrete to build that case around. More importantly, it helps decide what Rootmail becomes. You don’t need to be a developer. You don’t need polished feedback. And you definitely don’t need to understand email infrastructure. I’m looking for a small group of people willing to use something early, push on it, and help shape what comes next. If that sounds like you: Join the Rootmail beta → rootmail.io/beta Or leave a comment telling me one thing you’d want Rootmail to help you send. Let’s start there.

Oh grok @bot is about to be the end game I have some questions tho: How the usage be counted now and do I have to have subscriptions with each model to be able to use them?

I think one of the biggest mistakes you can make while building something is staying invisible until it is “ready.” I used to think: build → perfect it → launch → market it Now I think it looks more like: build ↓ share what you’re learning ↓ talk to people ↓ notice what resonates ↓ build better ↓ share again Marketing shouldn’t begin after the product is finished. The things you learn while building are already content. The architecture decision that took you 3 hours to figure out. The feature you built and later deleted. The assumption users proved wrong. The tiny improvement that suddenly made the product feel obvious. The bug that taught you how the system actually works. Those things may feel ordinary when you are inside the work. But to somebody else, they are useful. And over time, something interesting happens: people don’t just discover the product. They understand how you think. They watch it evolve. They start rooting for it. And some of them eventually become users. I’m trying to do much more of this with what I’m building now. Less: “Here is my product. Please try it.” More: “Here is what I’m trying to build, what I’m learning, what went wrong, and where I’m going next.” Build in public doesn’t have to mean sharing everything. It just means not hiding the journey that could have helped someone else. What’s something you learned while building recently that would probably make a good post?

I’m 9 followers away from 200😮‍💨🫡 Lets connect 🤝

Google just introduced Gemini 4 Argon — its new frontier model, and the first model in the Gemini 4 generation. What stands out to me is that Google isn't positioning Argon as simply a better chatbot. It is being built for deep reasoning across complex, long-running workflows: software engineering, finance, legal work, computer use and cybersecurity. And Google is already using it internally. Argon has been applied to debugging, large codebase migrations and infrastructure optimization — the kind of work where an AI has to maintain context, use tools, respond to failures and keep going rather than generate one good answer and stop. That distinction matters. We are moving from: Ask → Think → Answer toward: Goal → Investigate → Plan → Act → Test → Revise → Continue The unit of interaction is slowly becoming less about the prompt and more about the job. The cyber side is especially interesting. Google says Argon can autonomously find, validate and patch critical software vulnerabilities. And rather than immediately putting that capability into everyone's hands, Google is starting with selected cyber defenders and trusted testers through its Fairwind Program before expanding access more broadly. That may tell us as much about where frontier AI is heading as the benchmark numbers do. As models become capable of acting for longer and touching more consequential systems, intelligence alone isn't enough. You also need: permissions sandboxing verification monitoring human escalation The next frontier isn't simply building a model smart enough to start a difficult task. It is building one reliable enough to finish it. Gemini 4 Argon feels like Google's attempt to push directly into that frontier. The AI race is becoming less about who has the smartest chatbot and more about who can turn intelligence into reliable, useful and controllable action.

Google announced Gemini 4 Argon, the first model in its new Gemini 4 generation. It is designed around software engineering, cybersecurity, legal and financial work, computer use, and the increasingly important category that connects all of them

For a day now, the algorithm has really enabled my account to reach people a bit more as I’m close now to 200 follow🙏 Let’s connect 🤝

Liking a post that says @bot or grok bot triggers an animation

Liking a post that says @bot or grok bot triggers an animation

For the last few years, almost every major AI launch has invited us to ask the same question: How intelligent is the new model? OpenAI’s DevDay 2026 made that question feel slightly outdated.

Eventually, I think America.gov needs an app

Where did Astra go for plus users?🤔

I just got my x number

OpenAI DevDay 2026 feels less like “here are some new models” and more like OpenAI showing us what it thinks the next era of software looks like. The big idea: AI is moving from something you talk to → something that actually works with you. A few of the biggest announcements: • Dots — always-on AI agents that can keep working toward a goal, use tools, operate across apps, and run with their own permissions and guardrails. • ChatGPT Space — a shared workspace where people, ChatGPT, and agents can collaborate across documents, tasks, presentations, spreadsheets, plugins, and more. • GPT-6.1 Sol — positioned as a more economical model for serious agent workloads, where hundreds or thousands of model calls can happen behind one task. • Ultrafast inference — much faster execution for workloads like coding and long-running agents, where latency compounds quickly. • A higher-end ChatGPT tier — aimed at users pushing significantly heavier Codex, agent, and compute workloads. • Agents infrastructure — OpenAI is increasingly taking care of things developers previously had to build themselves: state, orchestration, tool use, sandboxes, long-running sessions, multi-agent workflows, and more. And this might be the most important part. The progression looks something like: 2023: Here’s an intelligent model. 2024–25: Here are reasoning models, tools, apps, and agents. 2026: Give the AI a goal, a computer, your tools, context, permissions—and let it work. That changes the value stack considerably. The interesting question for developers is no longer only: “What can the model generate?” It’s becoming: “What entire job or workflow can I hand to it?” For startups, that could be even more significant. A lot of what previously required queues, orchestration logic, agent memory, retries, sandbox environments, tool routing, and state management is slowly becoming infrastructure you can just build on top of. Models are becoming one layer. The agent runtime may become the bigger one. And if that’s where OpenAI is heading, DevDay 2026 may end up being remembered less for a particular model launch and more as the moment the company started seriously pitching AI workers as a platform. There are still more DevDay sessions and announcements to unpack, but that’s the story I’m watching most closely.

This is actually pretty sick It’s like grok bot

Nice work team Here's my two cents: Eventually, America.gov should have an app