• john garrish

Slop-pocalypse

From Andrej Karpathy, former Director of AI at Tesla and a founding member of OpenAI:

“Given the latest lift in LLM coding capability, like many others I rapidly went from about 80% manual+autocomplete coding and 20% agents in November to 80% agent coding and 20% edits+touchups in December. i.e. I really am mostly programming in English now, a bit sheepishly telling the LLM what code to write… in words.” https://x.com/karpathy/status/2015883857489522876?s=20

Aakash Gupta always has a good insight:

“Engineer” as a job title is splitting into two completely different professions: people who orchestrate agents and people who manually write code. The pay gap between those two is going to get very large, very fast.” https://x.com/aakashgupta/status/2015899420592947669?s=20

Author’s note:

I think they need another P in the term - because I want to pronounce it Slop (hard p) then POCalypse. Not, sloppalypse which I can barely say. Say it like Snow-pocalypse. Ok, now I’m just being picky.

What this all means for PMs

No matter how cranky developers are about PMs, the reality I’ve seen over and over is - PMs are routinely disregarded in many software organizations because they often don’t speak at a level of technical depth that Devs are comfortable with. He who owns the compiler wins, and unless you have strong strong strong executive support, PMs often chase around trying to speak the language of Dev.

To me - having PMs focused on the market is not a bug, it’s a feature. PMs should NOT work at a heavily technical level - they should always be honing “customer understanding” skills - what is valuable, for whom, and why.

Here’s what’s changing beneath our feet: AI tooling bridging this gap. In a world where code was just beyond the reach of “civilians” (ie the untrained programmer, ie a PM), developers held the keys to the kingdom. But as the engine cover comes off, and we now have tools that can peer inside the engine compartment and “speak the language of dev”, the landscape is shifting.

The shift means more people can now bridge the gap between “English” and “code.” And that includes PMs who are willing to learn. Or, perhaps vice versa - devs who want to move “up the stack” so to speak.

What’s not clear to me is -

  • In terms of job role, in both PM and Dev, is it the case that … PMs that can now provide more technical context, or Devs than can understand more business context? or both? or is it a new role entirely… AI PM? or ??
  • What are the inputs and outputs, in a world where coding is agentified, and maybe every part of the development lifecycle is agentified? Is the classical PRD obsolete?
  • What are the tools for the job? Are we all moving to Claude Code, or Cursor, or Gemini Studio, or some yet to be understood combination? What do we make of Jira and Confluence, and the mass of PPTs, DOCX etc etc that are floating out there (the typical tools up until now)?

Here’s one issue that I see, from where I sit - which is mostly on the PM side - the PMs who have made their living moving things from XLS to Jira? The PM-as-a-Waiter? That work is getting automated. Pretty soon you are going to need to pick a lane - either more into the technology stack, or more into the market. To the extent that AI automates away the low value work, then “cutting and pasting” will be the first to go. You can already see this with the various MCPs.

So as a PM - the choice is - either you dig in, and begin to bridge that gap between “English” and “code”… or you go somewhere else in the problem set.

My POV is - the best way is up to the customer, or maybe out to the market would be another way to say it. If you are a PM who has made Jira tickets your life, now is the time to build skills. Change your worldview. Use your superpowers in new ways. How do you understand what customers need, and why, and how to rank, sort, filter and articulate all of this ambiguity into revenue-bearing problems to solve? With the expectation that you are finding time for these activities by lowering your admin work via AI automation.

AI automation is the new normal for PM. It’s upon us. The good news is that the tools that are disrupting dev, are also disrupting PM. For example, if you haven’t taken Carl Velotti’s Cursor for PM course, do it now, today, don’t wait. It is an eye-opener.

The other topic is Product Marketing - I think the world of PMM is going to be impacted even more, simply because most of the problems they tackle are quite literally language problems. A post for another time.

Copy to .md - I’m seeing this on more pages

Copy from Claude into LLM friendly text - love this. I added this to my site (this site) for “code” blocks. But not the whole page. Maybe I need to add for the whole page?

Claude copy page dropdown

That’s all for now, folks. Stay warm, stay safe. -j