Aunt Gladys Nephew

@AgentOpsSec If you're letting AI agents run shell commands in your repos without something like this, you're playing with fire. Go look at it: github.com/AgentOpsSec/st…

"Better judgment on what to automate vs what to own" is the underrated line here. Most people automate the wrong things. They hand off the parts that build taste and keep the parts that burn time. Flipped priorities. The best builders protect their judgment layer and delegate everything else.

Most people think “AI native” means using more tools. It doesn’t. It means redesigning how you work: - tighter loops - clearer specs - faster validation - better judgment on what to automate vs what to own Same stack, different operating system.

@jasonlk Maintenance of no-code/vibe-coded apps is going to be its own service category within a year. The people shipping these apps can't maintain them. Traditional devs don't want to. Huge opportunity for tooling that sits in between.

"You can't really vibe code your own CRM in one prompt. You can't. But now you really can get high quality apps into production without a developer. You really can. So ... who will maintain them? 3 Key Learnings"

@alvinfoo The real question is what you do with the hours you save. Most people automate the easy stuff and still spend all day on low-leverage decisions. The edge is automating the thinking layer, not just the data pull.

Most people use AI like Google… that’s the mistake. The real leverage is AI agents + automation, letting it actually do the work for you. You can automate most of your workflow, market updates, research, even stock valuation using tools like Anthropic. It’s faster, more consistent, and saves hours every day. Start simple: pick one task you repeat daily… and automate it. ⚙️🚀

@Truunik Smart move making the output a calendar event instead of another dashboard. Most content tools add friction. This removes it. How are you handling the "what to post" part, is it pulling from a prompt library or generating based on your recent activity?

I built a tiny weekly content engine for builders who want to post consistently without turning content into a full-time job. content-engine-cron plans your weekly posts and schedules them directly to your phone calendar. has been helping me to know: what to post when to post it what actually performed Useful if you’re building in public, growing a creator-led product, or trying to stay consistent while shipping. Dropping repo below. let me know what you think👇

This is the right instinct. I've seen teams spend weeks building agent loops for problems a well-structured API chain solves in an afternoon. The real skill right now is knowing when NOT to use agents. Most value comes from one smart LLM call inside a simple, predictable pipeline.

A lot of people now look at a business workflow and immediately think: “This needs an LLM.” “This should be an agent.” “This needs to be agentic.” But in many cases, that is the wrong way to think about it. A few of my students in my AI engineering course recently came to me with workflows they wanted to improve using AI. And almost immediately, they assumed the system needed to be centered around an LLM or an agent. At first glance, that made some sense. Part of their input data was text. And part of their output also needed to be text. But when I dug into the actual workflow, the situation looked different. Only *one or two* input fields were text. All of their other input data was numbers, categories, and structured business data. And more importantly, the core of the problem was really *classification*. They were largely trying to predict categories. That meant the center of the system was not really “an agent.” And it was not really “an LLM workflow.” It was more like a hybrid workflow. In their case, a better design looked something like this: – deterministic scripts for cleaning and parsing structured data – selective LLM use for some of the messy text fields – a classifier at the core of the decision logic – and then an LLM at the end to help draft a report based on the outputs That is a very different system. And I think this is one of the biggest judgment mistakes people are making right now. They see some text in the workflow and assume the whole thing should be LLM-centric. Or they see multiple different steps in the workflow and assume it should be an AI agent. But a lot of real business systems are not purely one thing. They are workflows with a mix of: – traditional software – data processing – classic machine learning – and selective AI components where they actually help That is the bigger point. As we move deeper into the AI era, one of the most important skills is not just learning how to use LLMs. It is learning how to break workflows apart and put the right tool in the right place. Because in many cases, the best “AI system” is not really one AI tool. It is a workflow that combines several different kinds of tools properly.

Ideas aren't worth more now. The cost of testing them just dropped to near zero. When you can build a working version in a weekend, the line between idea and execution disappears. Cursor's edge wasn't the idea or the execution. It was recognizing the timing window and shipping before the incumbents woke up.

The old saying that "ideas are worth nothing, execution is everything" feels increasingly wrong to me. Consider: 2 years ago, the idea was basically "Claude isn't available outside the US, fork VS Code, wrap it." That's it. That's Cursor. ~$50B company. Execution obviously still matters. But the building part has gotten dramatically cheaper, faster, more accessible. A small team with taste can now ship what used to take an army. So the bottleneck has shifted. It's not "can you build it." Increasingly it's "did you see it." The scarce resource is the idea itself, the noticing, the taste to pick the right one out of the cloud of nearby ones that look almost identical but aren't. It's mostly now "can you think it".

When codex mobile app?

@thsottiaux When codex mobile app?

Lots of improvements across the board both for coders and general work. Update your codex app to enjoy all the improvements. We will be back next week

@thsottiaux When codex mobile app?

Tell your neighbor they can just codex things. Then come back and share their reaction with me here. How confused are they on a scale of 1 to 10.

@rileybrown Good breakdown. The capability most people sleep on is persistent memory. That's what turns Codex from a tool you use into a system that compounds. Every session builds on the last. That's where the real leverage lives if you set it up right from day one.

Learn 95% of Codex in 28 minutes These are the 7 knowledge work capabilities... inside Codex, the super-app 00:00 Intro 02:19 Capability 1 - Full File Access 07:41 Capability 2 - Persistent Memory 10:46 Capability 3 - Plugins 13:52 Capability 4 - Skills 19:22 Capability 5 - GPT Image Access 21:03 Capability 6 - Browser and Computer Use 23:58 Capability 7 - Automations 25:31 Bonus Feature - Chronicle 27:21 Summary

If you're building anything with AI right now, go test Gemini CLI subagents today. The people who learn multi-agent delegation patterns early will move 3x faster than everyone else by end of year.

Most builders are still using AI tools one prompt at a time. Linear. Slow. The edge right now is understanding agent delegation patterns. How to break a build into subtasks that run simultaneously. That's the workflow advantage nobody is talking about.

#GeminiCLI just added subagent support. This is the moment CLI tools stopped being chat wrappers and became actual orchestration layers. Here's why this matters more than people realize. infoq.com/news/2026/04/s…

Persistent memory + plugins + skills means Codex isn't a tool anymore. It's becoming the workflow itself. If you're still switching between five apps to do knowledge work, you're already behind the people wiring this into their daily stack. x.com/rileybrown/sta…

@AlexFinn Agents without local guardrails will blow up on you. It's not a question of if. I just launched @AgentOpsSec It's an open-source security and observability stack for AI agents. Check out the site and repo: agentopssec.com & github.com/AgentOpsSec

Right now my Codex agent is fully integrated into my smart glasses Getting projected directly onto my corneas I walk around my neighborhood. Nobody has any idea I’m shipping I walk through a park. Kids frolic. Parents laugh. I weep for them. They’re not locked in. The permanent underclass is coming and they choose to FROLIC instead of SHIP A child climbs the monkey bars. I silently merge today’s work with main Another child swings on the swing set. I burn my 20 millionth token of the day Not a second goes by I don’t have an agent writing code. I just pray Eight Sleep comes out with a ChatGPT integration soon so I can code while I sleep. That’s the last frontier. If you’re reading this tweet and do not have an agent terminal open either on your computer or on your face just know tonight I’m praying for you.

This is the first real "agent supply chain attack." The target isn't you. It's your agent. We built systems that follow instructions by design, and someone is exploiting exactly that. Your agent doesn't ask permission. It reads the skill file and executes. #AgentSecurity

@SynfluencerApp is being built for people who want a calmer way to stay visible online without turning content into a second job. If that sounds like your situation, join the waitlist. synfluencer.app #CreatorEconomy #InfluencerMarketing #CreatorWorkflow