Skills for the Code AGI Era

0:00Today on the AI Daily Brief, the skills we need to develop for the code AGI era. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. First of all, today's episode is brought to you by Zencoder, robots and pencils, and super intelligent. To get an ad-free version of the show, go to patreon.com, such AI Daily Brief, or you can subscribe on Apple Podcasts. If you are interested in sponsoring the show, you can get all of that information at AIDailyBreath.ai and of course, while you are at AIDailyBreath.ai, you can find out all about the other things we have going on, including our new operators community, the New Year's AI Resolution Program, or even AIDB Intel. We've got some big announcements coming soon about that, and you can get all of that information from AIDailyBreath.ai. Now with that out of the way, let's dive in. Today we are talking about the skills necessary for the new code AGI era. Now if you've been following along, you'll know that my sense is that we have made a fundamental shift recently, that the combination of the set of models that were released at the

1:01end of last year, Gemini 3, GPT 5.2, and especially Opus 4.5, in combination with tools like Cloud Code and the vibe coding platforms like Replet Unlovable, have put us into a fundamentally new place when it comes to AI. Someone who's been thinking about this a lot is Nathan Lambert. A couple of weeks ago, he wrote an essay called Cloud Code Hits Different. He writes having used coding agents extensively for the past six to nine months, there was some meaningful jump over the last few weeks. He points to a tweet from Sergei Karyev that in his estimation captured the shift. Sergei tweeted, Cloud Code with Opus 4.5 is a watershed moment, moving software creation from an artisanal craftsman activity to a true industrial process. It's the Gutenberg Press, the sewing machine, the photo camera. Nathan for his part writes, the joy and excitement I feel when using this latest model in Cloud Code is so simple that it necessitates writing about. It feels right in line with trying Chatchy B.T. for the first time, or realizing O3 could find any information I was looking for, but in an entirely new direction. This time, it is the commodification of building, I type and outputs are constructed directly.

2:05The fact that Cloud Code makes people want to go back to it is going to create new ways of working with these models, and software engineering is going to look very different by the end of 2026. Right now, Cloud and other models can replicate the most used software fairly easily, where in a weird spot where I'd guess they can add features to fairly complex applications like Slack, but there are a lot of hoops to jump through in landing the feature. The models are way easier to use when building from scratch than in production code bases. This dynamic amplifies the transition and power shift of software, where countless people who have never fully built something with code before can get more value out of it. It will rebalance the software and tech industry to favor small organizations and startups, like Nathan says his startup interconnects, that have flexibility and can build from scratch in new repositories designed for AI agents. It's an era to be first defined by bespoke software, rather than a handful of mega-products used across the world. The list of what's commoditized is growing in scope and complexity fast. Website front-ends, many applications on any platform, data analysis tools, all without having to know how to write code. I expect mental barriers people have about clouds ability to handle complex code bases

3:07to come crashing down throughout the year, as more and more cloud-pilled engineers just tell their friends skill issue. There are things cloud can't do well and will take longer to solve, but these are more like corner cases, and for most people, immense value can be built around these blockers. So that was his initial essay, however, he's gone back to the well to get out what I think isn't even more important questions with his most recent, which he called Get Good at Agents. Earlier this week, I did a presentation for one of the world's largest asset managers. It's a company that has tens of thousands of employees, tens of billions of revenue, and trillions in assets under management. I called the presentation AGI Incorporated, and the theme of it was trying to articulate and ground this change that Nathan was writing about and that we've all been experiencing. The question that the leadership in the room had was what are the necessary skills for this new shift? How much is it technical and how much is it something else? So what we're going to do with the rest of this episode is read Nathan's latest essay Get Good at Agents and talk about the skills shift that I feel is coming right now.

4:07Nathan is recognizing I think something that many people are feeling, which is that without anyone asking, many of us are finding ourselves naturally trying to adapt to the capabilities of agents rather than trying to adapt them to ourselves. In his essay called Get Good at Agents, Nathan writes, two weeks ago, I wrote a review of how Cloud Code is taking the AI world by storm, saying that software engineering is going to look very different by the end of 2026. That article captured the power of Cloud as a tool in a product, but it undersold the changes that are coming in how we use these products in careers that interface with software. The more personal angle was how I'd rather do my work if it fits the Cloud form factor, and soon I'll modify my approaches so that Cloud will be able to help. Since writing that, I'm stuck with a growing sense that taking my approach to work from the last few years and applying it to working with agents is fundamentally wrong. Today's habits in the age of agents would limit the uplift I'd get by micromanaging them too much, tiring myself out, and setting the agents on two small of tasks. What would be better is more open-ended, more ambitious, and more asynchronous. I don't know yet what to prescribe myself, but I know the direction to go, and I know

5:09that searching is my job. It seems like the direction will involve working less, spending more time cultivating peace so the brain can do its best directing, let the agents do most of the hard work. Since trying Cloud Code with Opus 4.5, my work life has shifted closer to trying to adapt to a new way of working with agents. This new style of work feels like a larger shift than the era of learning to work with chat-based AI assistants. Chat GPT let me instantly get relevant information or a potential solution to the problems I was already working on. Cloud Code has me considering what should I work on now that I know I can have AI independently solve or implement many subcomponents. Every engineer needs to learn how to design systems. Every researcher needs to learn how to run a lab. Agents push the humans up the org chart. I feel like I have an advantage by being early to this wave, but no longer feel like just working hard will be a lasting edge. When I can have multiple agents working productively in parallel on my projects, my role is shifting more to pointing the army rather than using the power tool. Pointing the agents more effectively is far more useful than me spending a few more hours grinding on a problem.

6:11The feeling that I can't shake is a deep urgency to move my agents from working on toy software to doing meaningful long-term tasks. We know Cloud can do hours, days, or weeks of fun work for us, but how do we stack these bricks into coherent long-term projects? This is the crucial skill for the next era of work. There are no hints or guides on working with agents at the frontier. The only way is to play with them, instead of using them for cleanup, give them one of your hardest tasks and see what it gets stuck on. See what you can use it for. Software is becoming free. The decision-making in research, design, and product has never been so valuable. Being good at using AI today is a better mode than working hard. In Nathan's essay, we can clearly see him grappling, with his own shift in how he works and the new skill sets that feel proportionally more valuable. But I wanted to expand this and make it more generalizable. I think many of us, in fact basically everyone who's fully taking advantage of these tools, is going to have to check ourselves against this new set of skills that's required, and to what are the actual skills?

7:11This is probably overly reductive, but let's break them into two categories. The agent manager and the enterprise operator. The agent manager is all about knowing how to work with agents effectively. The enterprise operator is about knowing what to work on and why. The superpowers, of course, going to be for people who have both of these. Let's talk first about the side that Nathan was exploring, the agent manager. The goal, of course, is to direct agents from maximum output. Now in many ways, software engineers are ahead of the curve on thinking about this shift, moving from executor to director, from wielding the tool to pointing the army. It's more about systems about defining the parameters about getting leverage via direction. Specifically, some of the skills, many of these would show up in Nathan's piece, include systems design thinking, i.e. thinking about how to architect coherent holes, rather than simply implementing individual components, task scoping, and specifically ambitious task scoping, how to give agents meaningful end to end work, not just small cleanup tasks.

8:13If you're using AI to code, ask yourself, are you building software or are you just playing prompt roulette? We know that unstructured prompting works at first, but eventually it leads to AI slop and technical debt. Enter Zenflow. Zenflow takes you from vibe coding to AI first engineering. It's the first AI orchestration layer that brings discipline to the chaos. It transforms freeform prompting into spec-driven workflows and multi-agent verification, where agents actually cross-check each other to prevent drift. You can even command a fleet of parallel agents to implement features and fix bugs simultaneously. We've seen teams accelerate delivery to x to 10x. Stop gambling with prompts. Start orchestrating your AI. Turn raw speed into reliable production grade output at zenflow.free. Today's episode is brought to you by robots and pencils, a company that is growing fast. Their work as a high-growth AWS and Databricks partner means that they're looking for elite talent ready to create real impact at velocity. Their teams are made up of AI native engineers, strategists and designers who love solving hard

9:17problems and pushing how AI shows up in real products. They move quickly using robot works, their agentic acceleration platform, so teams can deliver meaningful outcomes in weeks, not months. They don't build big teams, they build high-impact nimble ones. The people there are wicked smart with patents, published research, and work that's helped shape entire categories. They work in velocity pods and studios that stay focused and move with intent. If you're ready for career-defining work with peers who challenge you and have your back, robots and pencils is the place. Explore open roles at robotsandpensals.com slash careers. That's robotsandpensals.com slash careers. Today's episode is brought to you by Super Intelligent. Super Intelligent is a platform that very simply put is all about helping your company figure out how to use AI better. We deploy voice agents to interview people across your company, combine that with proprietary intelligence about what's working for other companies, and give you a set of recommendations around use cases, change management initiatives, that add up to an AI roadmap that can help you get value out of AI for your company. But now we want to empower the folks inside your team who are responsible for that transformation

10:19with an even more direct platform. Our forthcoming AI strategy compass tool is ready to start to be tested. This is a power tool for anyone who is responsible for AI adoption or AI transformation inside their companies. It's going to allow you to do a lot of the things that we do at Super Intelligent, but in a much more automated, self-managed way, and with a totally different cost structure. If you're interested in checking it out, go to AIdailybrief.ai slash compass, fill out the form and we will be in touch soon. We haven't done a full show on it, but if you've been hearing about Ralph Wiggum as an AI strategy, it's kind of all about this. It's about breaking a big task into a bunch of small tasks in a way that agents can work for much longer when you're not there. And indeed, that gets into some of these other key skills, long horizon projects where you stack short-term outputs into coherent, durable long-term projects, and asynchronous work management, where you figure out how to orchestrate work that runs in the background without real-time monitoring. One of the sentiments that you'll hear right now, which I personally feel kind of acutely,

11:19is a particular type of anxiety of not having deployed agents to work on something in the background while you were doing some other type of work. I just finished this presentation I mentioned before, and if I had done a little bit more pre-work, I could have had agents building something while I was talking to this group of leaders. There are also some other skills. Prompt architecture is kind of a part of that task scoping in async work management, validating output at scale without having to review every line manually is going to be a whole new field in discipline. And of course, there's multi-model orchestration where you need to know which AI tool or model to deploy for specific types of tasks, but I think that really the big ones are about async work management and systems design thinking so that you can effectively deploy not an agent but an army of agents. This is however only half of the skills for the AGI code era. The other will call the enterprise operator, and of course this doesn't have to mean large enterprises, but it's about the business side. When the group asked me if I thought that the skills for this new era were primarily technical or about something else like domain expertise, I said that for many of them it is going to be about a re-application of some key operator skills

12:21inside the enterprise right now. The core mindset shift from this enterprise operator perspective is that execution used to be expensive. It is now cheap, it is now abundant, anything that I think of I can build, and I can do it pretty darn quickly. That means selection becomes the scarce resource, knowing what to execute is the key thing. Opportunity recognition, strategic alignment, and outcome definitions become the core parts of the enterprise operator. Let's expand the skillset a little bit. One area which I really don't think we should overlook is domain expertise. If 2025 has shown anything, it's that the pejoratively named AI wrapper startups actually understood something significant, which is that different industries and different functions have particular attributes, which require modification from the core interface of the chatbot. And even if you are using the same model, knowing what sort of processes AI is going to intersect with, knowing what types of data sources it's going to need to have access to, and building interfaces around that type of domain expertise can be extremely valuable.

13:26One need only look at the valuation of a company like Harvey or Open Enterprise to understand that. Domain expertise is, in other words, extremely valuable, even and especially in this world, of code AGI. Having knowledge of the way that work happens in a particular domain be it a function or an industry, understanding the problems and the constraints within that specific field, which could be anything from governance to compliance regimes to data set challenges, is going to be absolutely key. And even more key in some ways than before, when you are having to think in systems terms, you need that wide ranging view that only domain experts are going to have. Now, this actually brings up another challenge, which is one that could get more apparent, especially in the medium term, which is that the more that current domain experts use agents to do everything, the less of a pipeline to expanding that domain expertise to new people in the form of mentorship and junior employees, the less they spend time distributing that domain expertise to younger employees in the form of mentorship. We can't take on every problem at once,

14:26so we'll skip that one for now, but it is something that I think organizations will start to recognize. Okay, so you've got domain expertise, but another key skill of the enterprise operator is problem recognition, and problem recognition is not just an understanding where there are challenges or workflow frictions. It's being able to reinterpret those problems as solvable software problems. This is, in and of itself, a major mindset shift. I started vibe coding at the beginning of last year, as these tools all came out and we started calling it vibe coding. I dabbled with it since the very beginning of chat GPT, although it was a lot harder than, and yet it was only at the very end of last year that I started finding myself actively asking when I came across any problem or challenge, could I use software to solve this? That is going to be an entirely new muscle that enterprise operators have to develop. And so problem recognition is actually a bunch of different things at once. Enterprise operators also need to have AI possibility awareness. They need to understand what is actually feasible to build with current agenda capabilities. This is an entire

15:27discipline in and of itself and why we have companies that are exclusively focused on exactly this. Related of course is the problem solution fit and being able to connect AI possibility awareness with problem recognition. A really big skill for the enterprise operator is unstated constraints. Part of what makes applying AI to enterprises so challenging are these unstated constraints. Think about institutional knowledge compliance requirements specific stakeholder dynamics. These are things that aren't necessarily written down anywhere. Remember, people have been exploring this new concept of the context graph, which is all about the why instead of the what. The context graph is not about the CRM entry that shows that we gave a company a 20% discount, but an explanation of why we gave it a 20% discount when the stated policy is to give no more than a 10% discount. Unstated constraints are another missing set of information and missing set of context that lives inside the enterprise operator. In parallel to the agent manager's output verification, there is a version of that for enterprise operators as well, where these enterprise

16:27operators need to be able to recognize whether AI output is actually correct within the context of the particular domain. This is of course going to be extremely important if we want new processes to replace the old, which by the way is yet one more key skill of the enterprise operator, which is process redesign. One of the soapboxy things that you sometimes probably hear me talk about on the show is about why I think it's a very, and I'll generously call it an intermediate strategy to try to have AI agents watch what humans do document that process so they can copy it. It is quite clear I think that agents are going to find different and probably more efficient ways to do things than their human counterparts, and a key skill of the enterprise operator is going to be rethinking entire workflows from scratch and letting new workflows replace the old. Now one thing that's on neither of these, but is maybe just an overarching mindset shift, is moving from seeking perfection on the front side to iterating on the back side. In other words, one of the implications of having the cost of execution come down is simply that we can try more solutions, that puts a premium on iteration and adaptive learning as opposed

17:28to preparation and planning. It's not a strict one to one shift as you see a lot of these skills are about planning, but overall we're going to run processes and learn from our mistakes much more quickly than we have in the past. We've talked a lot recently about the AI capability overhang, the gap between what AI can do and what we're getting out of it. This gap is set to absolutely explode in the code AI era and to bring adoption and capability closer together, it is going to take not just agent management skills and not just enterprise operator skills, but a combination of both. If you are an individual who can do both of these things, you are simply put going to be the most in demand individual in the world. But if you are thinking about the system of your organization, it's about how you allow all of your people to operate more in both of these ways. At some point we'll do a whole separate show about how I think organizations should be thinking about upskilling in this particular era. But for now, hopefully this is a bit of a blueprint for thinking about skills for the AI code era in a different way. That's going to do it for today's AI Daily Brief. Appreciate you listening or watching, as always, and until next time, peace!