The next article initially appeared on Wes McKinney’s weblog and is being republished right here with the creator’s permission.
Like lots of people, I’ve discovered that AI is horrible for my sleep schedule. Previously I’d get up briefly at 4:00 or 4:30 within the morning to have a sip of water or use the toilet; now I’ve hassle going again to sleep. I could possibly be doing issues. Earlier than I might get a strong 7–8 hours an evening; now I’m fortunate after I get 6. I’ve largely stopped preventing it: Now after I’m rolling round restlessly in mattress at 5:07am with concepts to feed my AI coding brokers, I simply stand up and begin my day.
Amongst my interior circle of engineering and information science associates, there may be a variety of dialogue about how lengthy our aggressive edge as people will final. Will having good concepts (and plenty of them) nonetheless matter because the brokers start having higher concepts themselves? The human-expert-in-the-loop feels important now to get good outcomes from the brokers, however how lengthy will that final till our wildest concepts will be became working, tasteful software program whereas we sleep? Will it’s a light obsolescence the place we fortunately hand off the reins or one thing else?
For now, I really feel wanted. I don’t describe the way in which I work now as “vibe coding” as this seems like a pejorative “immediate and chill” method of constructing AI slop software program initiatives. I’ve been constructing instruments like roborev to deliver rigor and steady supervision to my parallel agent classes, and to closely scrutinize the work that my brokers are doing. With this radical new method of working it’s arduous to not be contemplative about the way forward for software program engineering.
In all probability the e-book I’ve referenced probably the most in my profession is The Legendary Man-Month by Fred Brooks, whose now-famous Brooks’s legislation argues that “including manpower to a late software program venture makes it later.” These days I discover myself asking whether or not the teachings from this e-book are relevant on this new period of agentic improvement. Will a gifted developer orchestrating a swarm of AI brokers be capable of construct advanced software program quicker and higher, and can the short-term productiveness features result in long-term venture success? Or will we run into the identical bottlenecks—scope creep, architectural drift, and coordination overhead—which have plagued software program groups for many years?
Revisiting The Legendary Man-Month (TMMM)
Considered one of Brooks’s central arguments is that small groups of elite folks outperform massive groups of common ones, with one “chief surgeon” supported by specialists. This results in a excessive diploma of conceptual integrity concerning the system design, as if “one thoughts designed it, even when many individuals constructed it.”
Agentic engineering seems to amplify these issues, because the high quality of the software program being constructed is now solely nearly as good because the people within the loop curating and refining specs, saying sure or no to options, and taming pointless code and architectural complexity. One of many metaphors in TMMM is the “tar pit”: “Everybody can see the beasts struggling in it, and it seems like every one in every of them might simply free itself, however the tar holds all of them collectively.” Now, we’ve got a brand new “agentic tar pit” the place our parallel Claude Code classes and git worktrees are engaged in fight with the code bloat and incidental complexity generated by their digital colleagues. You possibly can systematically refactor, however invariably an agentic codebase will find yourself bigger and extra overwrought than something constructed by human hand. That is technical debt on an unprecedented scale, accrued at machine velocity.
In TMMM, Brooks noticed {that a} working program is perhaps 1/ninth the way in which to a programming product, one which has the required testing, documentation, and hardening in opposition to edge circumstances and is maintainable by somebody aside from its creator. Brokers are actually making the “working program” (or “appears-to-work” program, extra precisely) an excellent deal extra accessible, although many newly minted AI vibe coders clearly underestimate the work concerned with going from prototype to manufacturing.
These issues compound when contemplating the closely-related Conway’s legislation, which asserts that the structure of software program programs tends to resemble the organizations’ group or communication construction. What does that appear like when utilized to a digital “group” of brokers with no persistent reminiscence and no shared understanding of the system they’re constructing?
One other “large thought” from TMMM that has caught with folks is the n(n-1)/2 coordination drawback as groups scale. With agentic engineering, there are fewer people concerned, so the coordination drawback doesn’t disappear however reasonably adjustments form. Completely different agent classes might produce contradictory plans that people should reconcile. I’ll depart this agent orchestration query for an additional submit.
No silver bullet
“There isn’t any single improvement, in both expertise or administration approach, which by itself guarantees even one order-of-magnitude enchancment inside a decade in productiveness, in reliability, in simplicity.”
—“No Silver Bullet” (1986)
Brooks wrote a follow-up essay to TMMM to take a look at software program design by way of the lens of important complexity and unintended complexity. Important complexity is key to attaining your objective: For those who made the system any easier, it could fall wanting its drawback assertion. Unintended complexity is every little thing else imposed by our instruments and processes: programming languages, instruments, and the layer of design and documentation to make the system comprehensible by engineers.
Coding brokers are in all probability probably the most highly effective device ever created to deal with unintended complexity. To assume: I principally don’t write code anymore, and now write tons of code in a language (Go) I’ve by no means written by hand. There may be a variety of dialogue about whether or not IDEs are nonetheless going to be related in a yr or two, when perhaps all we want is a textual content editor to evaluate diffs. The productiveness features are huge, and I say this as somebody burning north of 10 billion tokens a month throughout Claude, Codex, and Gemini.
However Brooks’s “No Silver Bullet” argument predicts precisely the issue I’m experiencing in my agentic engineering: The unintended complexity is not any drawback in any respect anymore, however what’s left is the important complexity which was at all times the arduous half. Brokers can’t reliably inform the distinction. LLMs are extraordinary sample matchers skilled on the whole lot of humanity’s open supply software program, so whereas they’re sensible at coping with unintended complexity (refactor this code, write these checks, clear up this mess), they wrestle with the extra delicate important design issues, which frequently haven’t any precedent to sample match in opposition to. In addition they typically are likely to introduce pointless complexity, producing massive quantities of defensive boilerplate that’s hardly ever wanted in real-world use.
Put one other method, brokers are so good at attacking unintended complexity that they generate new unintended complexity that may get in the way in which of the important construction that you’re attempting to construct. With a few my new initiatives, roborev and msgvault, I’m already coping with this drawback as I start to succeed in the 100 KLOC mark and watch the brokers start to chase their very own tails and contextually choke on the bloated codebases they’ve generated. Sooner or later past that (the following 100 KLOC, or 200 KLOC) issues begin to collapse: Each new change has to hack by way of the code jungle created by prior brokers. Name it a “brownfield barrier.” At Posit we’ve got seen brokers wrestle rather more in 1 million-plus-line codebases akin to Positron, a VS Code fork. This appears to assist Brooks’s complexity scaling argument.
I might hesitate to position a guess on whether or not the current is a ceiling or a plateau. The fashions are clearly getting higher quick, and the issues I’m describing right here might look charmingly quaint in two years. However Brooks’s important/unintended distinction offers me some confidence that this isn’t simply concerning the present limitations of the expertise. Determining what to construct was the arduous half lengthy earlier than we had LLMs, and I don’t see how a flawless coding agent adjustments that.
Agentic scope creep
When producing code is free, realizing when to say “no” is your final protection.
With the price of producing code now converging to zero, there may be virtually nothing stopping brokers and their human taskmasters from pursuing all avenues that will have beforehand been value or time prohibitive. The temptation to spend your day prompting “and now are you able to simply…?” is overwhelming. However any new generated characteristic or subsystem, whereas low cost to create, just isn’t costless to keep up, check, debug, and purpose about sooner or later. What appears free now carries a future contextual burden for future agent classes, and every new bell or whistle turns into a brand new vector of brittleness or bugs that may hurt customers.
From this angle, constructing nice software program initiatives perhaps by no means was about how briskly you’ll be able to sort the code. We are able to “sort” 10x, perhaps 100x quicker with brokers than we might earlier than. However we nonetheless should make good design choices, say no to most product concepts, preserve conceptual integrity, and know when one thing is “completed.” Brokers are accelerating the “simple half” whereas paradoxically making the “arduous half” probably much more tough.
Agentic scope creep additionally appears to be actively destroying the open supply software program world. Now that the bar is decrease than ever for contributors to leap in and supply assist, initiatives are drowning in torrents of three,000-line “useful” PRs that add new options. As builders grow to be more and more hands-off and disengaged from the design and planning course of, the brokers’ runaway scope creep can get uncontrolled shortly. When the particular person submitting a pull request didn’t write or absolutely learn the code in it, there’s seemingly nobody concerned who’s really accountable for the design choices.
I’ve seen in my very own work on roborev and msgvault that brokers will suggest overwrought options to issues when a easy resolution would do exactly fantastic. It takes judgment to know when to intervene and methods to preserve the agent in verify.
Design and style as our final foothold
Brooks’s argument is that design expertise and good style are probably the most scarce sources, and now with brokers doing all the coding labor, I argue that these abilities matter extra now than ever. The bottleneck was by no means fingers on keyboards. Now with the brand new “Legendary Agent-Month,” we are able to fairly conclude that design, product scoping, and style stay the sensible constraints on delivering high-quality software program. The builders who thrive on this new agentic period gained’t be those who run probably the most parallel classes or burn probably the most tokens. They’ll be those who’re in a position to maintain their initiatives’ conceptual fashions of their thoughts, who’re shrewd about what to construct and what to go away out, and train style over the big quantity of output.
The Legendary Man-Month was revealed in 1975, greater than 50 years in the past. In that point, quite a bit has occurred: great progress in {hardware} efficiency, programming languages, improvement environments, cloud computing, and now massive language fashions. The instruments have modified, however the constraints are nonetheless the identical.
Perhaps I’m attempting to justify my very own continued relevance, however the actuality is extra advanced than that. Not all software program is created equal: CRUD enterprise productiveness apps aren’t the identical as databases and different vital programs software program. I believe the median software program consulting store is totally toast. However my thesis is extra about improvement work within the 1% tail of the distribution: issues inaccessible to most engineers. It will proceed to require knowledgeable people within the loop, even when they aren’t doing a lot or any guide coding. As one current adjoining instance, my pal Alex Lupsasca at OpenAI and his world-class physicist collaborators have been in a position to create a formulation of a tough physics drawback and arrive at an answer with AI’s assist. With out such specialists within the loop, it’s rather more doubtful whether or not LLMs would be capable of each pose the questions and give you the options.
For now, I’ll in all probability nonetheless be getting away from bed at 5am to feed and tame my brokers for the foreseeable future. The coding is simpler now, and actually extra enjoyable, and I can spend my time excited about what to construct reasonably than wrestling with the instruments and programs across the engineering course of.
Because of Martin Blais, Josh Bloom, Phillip Cloud, Jacques Nadeau, and Dan Shapiro for giving suggestions on drafts of this submit.
