My site: https://natebjones.com
Full Story: https://natesnewsletter.substack.com/p/executive-briefing-the-leveling-crisis?r=1z4sm5&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
_______________________
What's really happening with technical skills in the age of AI? The common story is that engineers need to code faster — but the reality is more complicated.
In this video, I share the inside scoop on the new technical skill tree that applies to everyone:
- Why Andre Karpathy feels behind and what that means for leaders
- How leverage shifted from writing code to orchestrating probabilistic systems
- What the four-level skill tree looks like from conditioning to compounding
- Where authority comes from when you delegate generation to LLMs
Chapters:
0:00 Introduction: Andre Karpathy Feels Behind
01:15 What Changed: The Phase Transition in Technical Leverage
02:24 The Old World: Authorship Equals Authority
04:39 What Broke: Control Is No Longer the Default
06:56 What Broke: Effort No Longer Maps to Output
07:50 What Broke: The Abstraction Stack Got Inverted
09:23 What Broke: The Old Boundaries Don't Make Sense
11:00 The Root Node: Separate Generation from Decisioning
13:00 Level 1: Conditioning (Intent, Context, Constraints)
15:30 Level 2: Authority (Verification, Provenance, Permissions)
18:00 Level 3: Workflows (Pipelines, Failure Modes, Observability)
20:38 Level 4: Compounding (Evals, Feedback Loops, Governance)
22:53 The Factorio Analogy: Building Automated Factories
25:00 Closing: Orchestrating Uncertainty Without Losing Authority
Organizations that recognize this as a workforce-wide challenge and build deliberate skill trees around separating generation from decisioning will realize 10X speedups, while those clinging to technical versus non-technical hierarchies will fall behind in 2026.
Subscribe for daily AI strategy and news. For deeper playbooks and analysis: https://natesnewsletter.substack.com/
📝 Transcript
Transcribing... This may take a few minutes.
Top Comments
@EightBitMeddler
I’ve never worked harder to stay exactly where I am
184 likes
@adrianchynoweth517
The definition of knowledge work is now "orchestrate probabilistic components while keeping authority." Bang on man!
35 likes
@Hitescape
I'm a 70 year old self taught software hack. Never officially trained in software engineering but none the less relentless until I can achieve my goals. Your insights have helped me appreciate how much education I missed out on and why I had such a hard time using AI. Over the last 6 months, your insights have helped me learn so much on how to use the tools like Codex and Claude to unlock my ideas and create things light years beyond my coding capabilities.
57 likes
@luisgdelafuente
As an engineer who has to work with many non-engineers, I can tell you that you are both right and wrong. Yes, there's a massive paradigm shift and change in tools, but no, it won't get any easier for most people. It's not and has never been about software engineering, but rather about understanding how things work and how to build models in your mind.
45 likes
@ClintBaxley
I swear I am feeling this exact thing right now. But it is also because of the people who are in my circle. We are all ahead of the curve. Like when I was a programmer for the web in 95. I was way ahead, but felt I was behind.
42 likes
@hedleyfurio
Excellent signal to noise ratio - I thought i would summarize , but on third viewing I had written down over 90 % of what you said.
14 likes
@andreylukyanov2179
I think Andrey Karpathy feels behind because he tries to use AI to solve real world problems and he sees that this is hard. In one of his recent interviews he mentioned that he uses vibecoding for fun, but for real world problems he works on he rarely goes beyond using AI to autocomplete his small scope code ideas. This exactly matches my observations with using coding agents - as the scope of the project gets larger and larger the more time you have to spend to keep AI "in-line" doing what you want it to do.
I would love to see more depth and examples in this channel to support your claims and reasoning. Without this your arguments lack substance and prove/say little, for me personally.
18 likes
@RealMohammadEmran
Hey Nate, I really love the deep insights you share daily with the rest of the world and I hope it doesn't stop. With the internet full of bullc**p, contents (also "free"!) like this make the platform meaningful. As a mid-level engineer working in the German automotive industry, I regularly watch/listen to your messages and the lessons resonate with my personal experience and I also share them with my colleagues. Respect for you man!
Just a side note specific for this video, throughout this video it kept coming to my mind, the messages in "this" video will be even "better" delivered, if there were some graphical contents. Just a personal opinion. Not criticizing, in any way, the great effort you already put into making the content.
11 likes
@arterial
I'd have been lost without this channel. Thank you, Nate
20 likes
@break-nil
What an impressive overview of the current state of the art for development and project management
5 likes
@Pulsonar
Around 5:00 like a time bomb he says “Blast radius” which is a very apt description of the devastating impact these radical new fundamentalist insurgent models and alien workflows are having on developers working lives.
4 likes
@voncolborn9437
I'm a retired 73 year old, retired software engineer. This is all very interesting, yet at the same time, it is just a description of what was layed out several decades ago by Carnegie Mellon University's SEI. The Software Capabilities Maturity Model (SCMM) for software engineering organizations. It lays out how an organization goes from chaos to repeatability to improvement level of operations. Nate, you are so correct in all that you've layed out here. Working with LLMs for software development is at the same time an amazing experiment and more frustrating than working with a new engineer wanting to prove their metal by showing you all that they know, at the expense of of the project.
With all that you've laid out in this brief video tells me two things: 1. Experienced engineers will not be replaced. There is too much discipline and experience required to manage the AI. 2. The training programs need to get the young engineers on the right track and show them how it needs to be done properly and companies better plan on continuing to hire entry-level engineers so that they can grow into the senior positions, just as it has been from the beginning. You don't get adults without them starting out as babies.
2 likes
@GlennEuloth
I appreciate you putting words to what's been going on inside my head for the last few months! Thanks!
2 likes
@dangayle
This is a really nice breakdown. I’ve been trying to explain to other developers that they’re no longer “developers”, they’re project managers. Project managers have always had the same issues of requirements, testing, validation, non-determinism, etc.
2 likes
@bsarel
I think I’ve watched this like 5 times to make sure I mentally capture all of it.
What an excellent encapsulation of the current chaos that we call “the present”