Roman Jaquez

This meta-prompting technique needs a formal name. I use it frequently in the middle of a (e.g., Claude Code) session: 1. ask the agent to create a prompt for another agent to do X 2. rewind the session back a step so that it doesn't remember creating the prompt 3. give the same agent the prompt it just wrote This trick bypasses the meta-awareness trap. If you try this without the rewind, the outcome is less likely to succeed. The agent will think things like, "this is the prompt I just generated" and might even just asked you a question about the prompt rather than assume you're telling it to proceed. One of the main benefits is that you can powerfully steer the existing (context-primed) agent via the meta-prompt without first having to build up the context of another agent to build the prompt. Does this have a name? If not, what to name it? Ghost Prompt?

7

6 Comments

😅😅😅 GitHub Actions workflow allowed code injection through pull request titles, resulting in malicious versions of the nx package being published to npm (containing code that scans the file system, collects credentials, and posts them to GitHub as a repo under the user's accounts) https://lnkd.in/dFZ4dj92

12

POV: Everyone’s losing their minds over Claude Code subagents and you’re just sitting there like 🧍‍♂️🤔 Me for the past 3 weeks: “Hey Claude, split this into 4 parallel tasks” Claude Code: “Sure thing” spawns 4 subagents Me: “Cool” Everyone: “OMG SUBAGENTS ARE REVOLUTIONARY! GAME CHANGER! PARALLEL AI AGENTS!” Me: “…wait, was that not supposed to work?” Apparently I’ve been using “cutting-edge multi-agent orchestration” the masses didn’t know was a thing, I thought I was just asking nicely for help with my refactoring. Next you’re gonna tell me that saying “think harder” to get better results is some kind of advanced prompt engineering technique and not just me being demanding. 😭 Currently wondering what other “undocumented features” I’ve been stumbling into by just talking to Claude Code like a normal human being. 📚 Resources for fellow accidental experts: Complete guides, examples, and templates at: 👉 https://lnkd.in/g2VnyxEy Includes: 🤖 Subagent best practices (casual → AI teams) 🧠 Thinking levels mastery (think → ultrathink) 🛠️ Ready-to-use examples & templates Turns out "accidentally advanced" is the best way to learn AI development 🎯 #ClaudeCode #AccidentallyAdvanced #TwilightZoneMoment #UltraThink

5

With the release of Xcode 26 and iOS26, It's time for some new material. IN the next few weeks, my course Complete guide to SwiftUI will be updated with some of the changes for iOS26 and iPadOS26. Here's a bonus lesson I made after I finished production on the Xcode26 update explaining safe area insets and the new ConcentricRectangle shape.

2

With the advent of powerful coding AI's, how important is it for a senior level engineer to know a language in and out now? I know the paradigms well - functional, procedural, object oriented, prototypical... do the specifics matter anymore? In fact, I wonder, are there languages more suited for AI's to work with? I would be very interested to see if there are any studies about this: What languages work best with AI coding assistants... and why? Lets say we find the answer there... and AI assistants converge on some new unified programming language that they know, but we don't (at least not in the population at large). Is that the future for engineers, where we engineer through things like UML diagrams, behavior descriptions, and outcomes? Perhaps. But honestly, I am here for it. I love engineering, no matter what tool I use to do it.

8

17 Comments

I’m gonna make a bold prediction based on the direction I’ve seen coding agents move lately: I think companies in the professional services game are about to do some solid growth. Tools like Claude Code have made it possible to develop tailored, custom software solutions quite rapidly in organizations and build them from the ground up. There are a lot of names for this, but I like to call it the problem-out approach rather than a solution-in approach. What I mean is, software vendors tend to try and stuff an (often over bloated) software suite into a problem rather than trying to grow the solution organically with the customer. I can’t call this a mistake, it’s just what we’ve had. Now, enter AI assisted coding paired with development experience and subject matter expertise. My friends, if a large company or (caugh, caugh) government agencies were to build their own tailored software in house with the customer riding shotgun we would see some significant transformation in the space. I subscribe to Lean Startup principles and the concept of “build, measure, learn, and iterate” has never been more applicable than now (I see you, Google Rapid Innovation and Testing team) I can see why some of these large software companies dropped on the market recently. If a client or customer can build a useful, tailored, and easily reconfigured replacement for something like salesforce internally why would they opt for the expensive version? Professional service companies that have a proven process of working with customers, like we have at Steampunk, Inc. with Design Intelligence, are critically well-positioned to deliver at this pace and be the catalyst for this transformation. I genuinely think this the direction we’re headed and I think now is the time to lean in; both from the customer side and the service side. Good hunting. RLTW #ai #professionalservices #tech #coding #government #contracting #govcon #consulting

7

1 Comment