Amazing Excerpts from Steve Yegge, Sourcegraph, on GenAI, coding assistants, and the Death of the Junior Developer (ETLS Vegas 2024)

I’m so thrilled that Steve Yegge , Cody guy at Sourcegraph , presented at ETLS Vegas in August, on GenAI, coding assistants, and the death of the junior developer.

These are some of the fantastic highlights from his talk.

Anyone who has studied Amazon or Google likely knows his work. He famously gave the world an inside account of Jeff Bezos’s “Thou Shalt Communicate Only by APIs” memo. This accidental public post even landed him on the front page of The Wall Street Journal.

I finally got to meet him two months ago, and he’s as hilarious, irreverent, and brilliant as you might expect.

One outcome of that conversation was another epic Yegge post called The Death of the Junior Developer, based on his observations from helping build a coding assistant at Sourcegraph.

His was one of the many talks that day that led John Willis to say, "Today at ETLS felt like the 2009 Velocity Conference '10 deploys/day at Flickr' moment." This is often referred to as the start of the DevOps movement — this was amazing to hear, since he was actually there!

Yegge said: "I haven't been on stage in 15 years and it'll probably be another 15 years before I get invited back because I'm gonna say something that's gonna get me in trouble. And I just finished my slides an hour ago.

(Hahaha.)

He spent most of the last 30 years at four companies; Geoworks, Amazon, Google, and then Grab (the amazing application used by millions of people in Southeast Asia. It's a combination of Uber, Uber Eats, PayPal, etc.). It was amazing to use during my visit to Singapore in 2022.

Coming out of retirement after COVID, he's on a quest to understand the hopes, dreams, and problems that enterprise technologists face.

Yegge: "I'm basically like the cable guy. I just go house to house. I even get barked at by security and legal teams, and I get to understand their problems and see the patterns.

"Why? Because I get to go into all of your houses. I mean, not your literal houses, thank God, but your place as a business and I get to see your architecture and your hopes and your dreams and your ambitions and your problems. And you all open up to me and it's pretty awesome."

Yegge observes that life in engineering isn't as easy and happy as it was for most of the last decade. He eventually gives that factor a name: engineering entitlement.

His explanation: we've all been ZIRPed — Zero Interest Rate Policy.

Yegge: "The $2 trillion economic stimulus package.... Everybody had free money and jobs, jobs, jobs. Software engineering jobs!"

In contrast to the ZIRP-fueled overspending on engineering frivolity, he describes the opposite: the austerity that Amazon was famous for in the early 2000s.

He talked about early days of Amazon being a paragon of the complete absence of engineering entitlement.

Yegge: "I actually got warned about Walmart's influence on Amazon in 1999... 10% of Amazon at the time was ex-Walmart senior executives... They brought the Walmart culture, and their culture was basically one message... Do not allow your engineers to become entitled. For reasons I probably explained before, [at Amazon], we were not very entitled... because Bezos took his entire playbook from Walmart.

"And it wasn't until I went to Google in 2005 that I finally understood what entitlement meant... Because they were for engineers by engineers. They were princesses. Everyone got a pony. It was awesome. And I became one of the most entitled people there. I mean, I'm not proud of it. It's a real phenomenon."

Haha. This is irreverent Yegge at his best. ??

Yegge describes what happened when he posted his amazing "Death of the Junior Developer" post — which he wrote after he and I shared our observations on how it affects the roles of both senior and junior developers.

In short, GenAI helps senior developers achieve so much more. It can automate easy problems, and help solve difficult problems.

But it also enables senior devs to do things that normally would have required help from someone else. Steve shared a story with me:

His Head of AI built something in two days using a coding assistant — that person said, "this would normally have been one of the two projects I'd give to a summer intern."

This is great for senior developers! But it was definitely not so great for that junior developer who was going to do that potentially rewarding summer internship — that job req never got created.

This shows how and why junior developers can be negatively impacted by what Dr. Matt Beane calls the "novice optional" problem. Performing surgeries has always required at least three hands — so a senior surgeon would be assisted by a junior surgeon.

But with the advent of surgical robots, a senior surgeon can do the work all by themself! And woe to most of the poor junior surgeons who never get enough practice time with the surgical robots.

And now this is happening with developers. Which is what Steve wrote about here:

In his talk, Yegge said: "My post divided people right down the middle... The post was based on a really simple premise... The premise was that within 18 to 24 months, all code is gonna be written by LLMs. A lot of people didn't like that."

"People called me because they were so angry at me. Why? They were angry because their kids just graduated with computer science degrees and they're like, you're poisoning the well."

He shared so many interesting insights — among them, that LLMs are taking over the low-level creative tasks, leaving humans to oversee and refine the output.

To paraphrase:

All senior devs need to do is supervise, like it's a kitchen. Your'e a cook, and you've just been elevated to a master chef — you now have a bunch of robots that can do the prep, cook and all the other work. All you have to do is turn it into a good meal.

"And that's why there's this shift towards senior contributors. Because if you're a bad cook and I give you a bunch of robots, you're just gonna make a big bad meal, right?"

Haha, but not exactly in the funny sort of way, right? (I think about my 3 kids, who are 16, 14, and 14. How do you prepare them for a world like this?)

One of the most memorable parts of his talk was this: "Programming is changing. And many of you, most of you are still stuck in like February! That's how far and how fast this changes!"

Steve describes what CHOP (Chat Oriented Programming) looks like, and who is and isn't using it. His observations:

• Senior developers are more likely to use AI tools like LLMs than junior developers.
• LLMs can accelerate coding speed, making developers by 5-10x.
• There is a wave of new tools coming in, because the needs go way beyond traditional IDE capabilities — this was also mentioned by Idan Gazit, Senior Director of GitHub Next
• Discovery of information within large codebases will become the most significant challenge — "you've got thousands, tens of thousands of repos, you got wikis, you've got issues and project trackers."
• Until context window is infinite (which is unlikely), coding assistants reduce down to a RAG program — "Earlier today, Nubank was ingesting what, 50 petabytes a day. How is your LLM going to process that?")

Steve posits that organizations that aren't using GenAI for coding are at grave risk.

• There is a massive difference in how companies approach generative AI: some are like science-fiction, building out 5-10 year plans of how they're going to use GenAI.
• Others are doing nothing, because they say they've got other priorities like "migrations, blah, blah, blah."
• "They're gonna die, right? This is serious. Coding is changing out from underneath us and they're struggling trying to figure out whether they they can even use it or not!" (Because of compliance, etc.)
• You must assume that great self-hosted LLMs are eventually going to coming — you need to learn how to use these tools before they arrive. Local models will negate all of the security objections, but don't waste this time.

"I guarantee you self-hosted models are coming. It will be accessible to all of you in a way that makes your security and legal teams fine with it."

"I know a lot of you have tried coding assistants... In February. They weren't very good back then. But the release of GPT4o and Claude-3.5 made all the difference."

Steve believes the key to coding assistants working well boils down to a RAG problem:

"Everything is not going to fit in the context window. So you need an intelligent system that can go in and index, produce indexes of all of your stuff: search indexes, knowledge graphs, LLMs, semantic indexes."

Despite Steve working on Cody, he says the most important thing is to use one!

• "Chat oriented programming (CHOP) is what programmers are going to do before from now until the agents actually work."
• Humans will always be in the loops: "you need a human being to supervise it, to push it along, even with the agents."
• "CHOP is not gonna go away. Not for, not for a good long time."
• Focus on getting your code indexed — it will become critical path. He believes indexing code is a joint effort between companies and their code indexing tool. "It has to be a joint effort because you all have really complicated systems if you hadn't noticed from the talks."

Thank you for giving this wonderful and important talk, Steve!

You can watch Steve's complete talk here (just register with your email address):

https://videos.itrevolution.com/watch/1002959965

I’m thrilled Steve will be joining me for a Q&A at ETLS Connect next week—hope to see you all there!

Details and registration here: https://itrevolution.com/product/etls-connect-october-2024/