My “Dwell with Tim” dialog with Steve Yegge this week was a kind of periods the place you might think about the viewers leaning ahead of their chairs. And on a couple of event, when Steve bought significantly colourful, I imagined them recoiling. Steve has all the time been one of the vital provocative thinkers in our business, going all the best way again to his legendary 2011 platform rant that leaked from inside Google. Nowadays he’s channeling his power into Fuel City, an open supply AI agent orchestrator, and right into a relentless marketing campaign to shake builders out of what he sees as a state of denial about the place coding is headed.
And sure, Fuel City is certainly named after the gasoline depot in Mad Max: Furiosa, and even incorporates a managing agent named after the Mayor. However Steve’s Fuel City is something however dystopic. If something, it’s joyous. That offers you a deep sense of who Steve is: He goes into the deepest, darkest a part of the forest, finds one thing scary, after which does his greatest to redeem it.
We coated lots of floor: the eight ranges of coder evolution, the addictive pull of multi-agent workflows, grief and denial within the developer group, the bitter lesson, and why style could be the final remaining aggressive benefit. Listed here are a number of the highlights.
Everybody will get a chief of workers
Steve’s “Eight Ranges of Coder Evolution” framework has taken on a lifetime of its personal since he printed it as a part of the Fuel City launch publish. The primary 4 ranges are about more and more subtle IDE use; ranges 5 by way of eight are about coding brokers. Right here is an infographic displaying the eight phases that I confirmed to anchor the beginning of our dialog. Observe that I created this slide with Nano Banana 2. It isn’t straight from Steve.
The important thing transition, Steve argues, occurs at degree 5: Your IDE goes away and also you by no means open it once more. As Steve described it, when you notice Claude Code can write items of your code, you begin assembling them like Lego. However whereas one agent is working, you’re sitting there bored, so that you hearth up one other one. And one other. Earlier than lengthy, you’ve bought six brokers operating in parallel, and one in all them is all the time completed and ready on your consideration.
Steve drew an analogy to Amazon VPs who had government assistant assist. These individuals had been successfully two individuals. They didn’t have to fret about whether or not the printer was jammed, so they might spend all their time centered on the actual issues. Fuel City, Steve argues, is topologically related: It’s going to show all people into one thing like an government with a chief of workers. “All of us have a chief of workers now,” he stated. “All people’s going to have the ability to spend their time extra productively on no matter they need to spend it on as an alternative of determining the place the printer is jammed.”
On March 26, be a part of Addy Osmani and Tim O’Reilly at AI Codecon: Software program Craftsmanship within the Age of AI, the place an all-star lineup of consultants will go deeper into orchestration, agent coordination, and the brand new expertise builders must construct wonderful software program that creates worth for all contributors. Join free right here.
The AI vampire
That is Steve Yegge, so he’s not simply going to provide the upside. His publish on “the AI Vampire” defined how AI-assisted productiveness creates an insidious new form of burnout, and I made positive to ask him about that.
The previous model of overwork was your organization piling duties on you till you broke, he informed us. The brand new model isn’t your boss asking you to work additional hours. It’s Claude saying, “Is there the rest you’d like me to do on this undertaking?” And also you say sure, sure, sure, as a result of it’s enjoyable, as a result of it’s productive, as a result of the AI is your buddy, not your employer.
However there’s a twist. The AI is fixing all the simple issues and leaving you with nothing however arduous ones. In our dialog, I stated it might really feel like your bike journey is all hills now, and Steve instantly related it to watching Jeff Bezos in conferences at Amazon. Individuals would carry him shows the place they’d already solved each simple drawback, so Bezos was simply getting the arduous stuff, all day lengthy. “Now this occurs to you,” Steve stated. “Everybody’s Jeff Bezos, everybody’s an entrepreneur. Everybody has an enormous military of staff now. And I’m telling you, it’s exhausting.”
Steve informed us he naps day by day now, typically twice a day, feeling drained by the relentless cognitive depth. These brokers don’t simply make it easier to work quicker; they essentially change what sort of work reaches your desk.
On the flawed facet of the bitter lesson
We spent an excellent stretch on Richard Sutton’s “bitter lesson.” Sutton noticed that uncooked computation constantly beats methods constructed on human-engineered construction. Steve treats it much less as a paper and extra as a day by day working precept. “Not a day goes by once I don’t take into consideration the bitter lesson as a math method,” he stated, “no less than 5 occasions a day.”
His sensible take a look at is straightforward: In the event you’re writing code that tries to make the AI smarter, by including heuristics, parsers, common expressions to deal with what a mannequin may deal with, you’re on the flawed facet of the bitter lesson. He watches even his personal Fuel City contributors make this error, reaching for just a little regex hack when they need to let a mannequin do the cognition. (Steve does admit that typically you do must present prebuilt code if it saves tokens.)
Sutton wrote concerning the bitter lesson within the context of coaching packages to beat people at chess and go, nevertheless it’s extra basic than that, much more basic than leaning into as we speak’s AI in the best way that Steve does. I shared my very own first encounter with the bitter lesson, again in 1993 when O’Reilly created GNN, the primary business net portal. Being publishers, we curated a catalog of the very best web sites. Then Yahoo! got down to listing all of them, proscribing curation to placing them into classes. Then Google did it algorithmically, making a customized curation for each search. We all know which method gained. The bitter lesson isn’t nearly AI; it’s a couple of recurring sample within the historical past of expertise the place scale and computation overwhelm hand-tuned options.
I nonetheless imagine now we have to guess in opposition to the bitter lesson, as a result of if we simply quit, there might be no place for people sooner or later data financial system. However now we have to do it understanding that we aren’t going to win within the conventional sense. We aren’t going to outrace AI. We’ve to study to journey it.
For anybody in a company setting, you’ll naturally need to match AI into your present workflows. The bitter lesson says you need to as an alternative determine what the AI can do by itself first, after which construct a brand new workflow round that. I described Steve’s complete method as wanting the bitter lesson within the face and saying, “I’m going to show AI free on the whole lot I can, after which determine the place the human suits within the loop.”
Code is a liquid
Steve hit peak Yegge mode when an viewers member requested why they need to depart their IDE. His response was, as common, quotable:
In the event you’re your code, then you definately’re in a System One race and also you’ve parked your automotive and opened the hood and also you’re wanting on the engine. You’ve slowed time to the purpose the place everyone seems to be racing previous you and also you’re a frozen statue. Code is a liquid. You spray it by way of hoses. You don’t freaking have a look at it….
Look, I get it. That is painful for individuals. That is tremendous painful. For me to say these items is painful for individuals to listen to them. As a result of what I’m saying is your job goes to vary.…And there’s nonetheless lots of denial on the market.
What’s the primary section of grief? The primary section of grief. The entire world is in it proper now, Tim. They’re in denial. Proper. They’re grieving for what goes away. We’re on the finish of an period. An age, a golden age, perhaps, the place we programmers, we’re writing all of the code. And it was fantastic for 30, 40, 50 years or no matter. That period is ending and persons are grieving due to it. And I really feel for them. I’ve bought empathy, proper. However I’m additionally shedding endurance as a result of it’s 2026 and that is an exponential curve, and we don’t have time to sit down round and really feel pity for ourselves.
He sees that grief in every single place, however he particularly known as out Hacker Information, which he described as the house of “the brand new Amish.”
Style is the moat
One other of the viewers questions was about whether or not firms with deep pockets have all of the leverage and there’s no room left for people. Steve’s reply was emphatic: completely not. Steve made a passionate argument that creativity outweighs capital within the AI period. He’s sure there might be corporations that waste hundreds of thousands of {dollars} of tokens constructing software program that by no means sees the sunshine of day, as a result of they’d no style, no good concepts, simply brute-force era with out route. In the meantime, an entrepreneur with open supply native inference fashions and an excellent GPU can construct one thing that issues, in the event that they know what individuals need.
“The whole lot goes to return right down to style,” Steve stated. “Corporations don’t have a bonus anymore. As an entrepreneur, I believe it is a golden alternative for individuals to make enormous, enormous impression.”
It’s mentors all the best way down
Somebody from the viewers requested Steve a couple of query that’s on everybody’s thoughts: If senior builders have gotten PMs and juniors are being changed by AI, the place will new seniors come from? His reply was a basic reframe, and an replace of what he wrote in “Revenge of the Junior Developer.”
He made the case that your most junior engineers aren’t who you suppose they’re anymore. They’re your product managers, your SDRs, your finance and gross sales people, all of these individuals all through your organization who at the moment are constructing issues with AI. Your former junior engineers are literally well-trained engineers who make excellent mentors for this new backside layer. And people juniors get mentored by seniors, and seniors by principals. It’s mentoring all the best way down.
Steve identified that this connects to one thing Matt Beane (who was in our viewers) has researched on expertise acquisition: You don’t study from somebody 40 ranges above you; you study from somebody one or two ranges forward. Steve’s suggestion for corporations is to arrange round this. Discover mentors inside your group who’re only a step or two forward of the place every individual is, and convey everybody together with empathy for what persons are going by way of.
We’re going to construct greater stuff
One other viewers member requested about analysis displaying that AI atrophies important pondering pathways. I couldn’t resist leaping in with one in all my favourite historic analogues. Socrates stated the identical factor concerning the written phrase, arguing that it was impairing individuals’s capability to recollect. And he was proper, we did lose the flexibility to recite large quantities of literature from reminiscence. However we gained greater than we misplaced. Issues change.
I additionally shared a Rilke poem that I really like, about Jacob wrestling with the angel: “What we battle with is so small, and once we win, it makes us small. What we wish is to be defeated decisively by successively better beings.” If AI is atrophying your pondering, it’s since you’re not wrestling with arduous sufficient issues. The actual alternative is to be pushed, stretched, and defeated by greater challenges, and are available away stronger from the battle.
Steve agreed: “We’re going to construct greater stuff. That’s what everybody’s anxious about. What’s going to occur? And the reply is we’re going to construct greater stuff and it’s going to be enjoyable.”
Watch the complete dialog right here. Steve Yegge’s Fuel City is at https://github.com/steveyegge/gastown. His weblog posts on “The AI Vampire,” the “Revenge of the Junior Developer,” and “Software program Survival 3.0” are important studying for anybody navigating this transition.
