I asked cracked engineers for their hottest takes on AI. They didn’t disappoint. Here’s the list.
Table of Contents
AGI Is Not Imminent
GPT-7 will not be AGI. AGI is not coming from ever-greater zero-shot reasoning models.
AI Is The Great Equalizer
In a weird sense, AI is the great equalizer; for example, you cannot hide. If you’re that talented, you can amass extraordinary leverage. Great engineers are harder to find but easier to distinguish.
Managing AGI Expectations
Sam is just driving the AGI narrative, which is fine for the broader ecosystem, but will cause downstream disappointment when a zero-shot reasoning model cannot act as a general intelligence agent.
Small Models Have Big Impact
Small (localized) LLMs + great data will be good enough for 90% of GDP shift. The world is running on spreadsheets.
AI Is An Amplifier
AI is an amplifier. Just like giving people money amplifies who they truly are, AI shows you how good you really are.
AI Leverage For Top Engineers
The superior your engineering knowledge is, the greater the leverage AI can bring. The difference in leverage between okay engineers and 10x engineers is not 1.5, or even 5, it’s probably 20x, 100x.
The Continual Learning Problem
There are still some very key problems with AI, the biggest of which is “sleep-time compute” (continual learning outside of context). This probably represents the biggest gap between AI and humans right now.
Go Beyond Memorized Facts
AI is VERY good at memorizing. If all you bring as an engineer is memorized facts, there’s no reason to hire you.
Infinite Context Is Far Away
The infinite context window / sleep time compute thing is pretty far away. Memorization / recalling the training data is not.
The Human Context Advantage
The biggest advantage you have as a human is understanding everything and only having to pay that cost once. Once I understand a codebase, I won’t have to be reprompted with ALL the context, whereas every time I start a new chat, I will need to give the AI all that context.
Delegating Memorization To AI
I use AI to replace the “memorizing” parts – specific syntax, etc. I still make the decisions, but I can delegate to AI because I know “whether the thing is possible or not.”
AI Works In Good Codebases
If you’re a developer and you think AI sucks at working in your codebase, you probably have a messy codebase.
AI Won’t Replace Engineers
AI isn’t going to fully replace software engineers because that would require everyone being able to describe exactly what they want in their software (i.e., writing code). This is the age-old problem of the requirements never being quite specific enough.
The Future for Junior Developers
Everyone thinks junior developers are in trouble, but I actually think if they have solid fundamentals and can use AI, they will have a better shot than a senior engineer who learns slower and claims “AI isn’t that good.” Companies are getting rid of the least efficient engineers, not the least senior ones.
Key Takeaways
Here’s a recap:
- AGI won’t come from zero-shot reasoning models alone—localized LLMs with quality data will drive most economic impact.
- AI amplifies existing talent rather than replacing it, creating massive leverage gaps between good and great engineers.
- Critical AI limitations remain, particularly “sleep-time compute” and persistent context understanding.
- Humans excel at one-time learning and decision-making, while AI handles memorization and syntax.
- Junior developers with strong fundamentals and AI skills may outperform resistant senior engineers.
- Poor codebases become obvious when AI struggles to work with them – software engineering jobs are safe because defining requirements remains fundamentally difficult
Conclusion
The engineers paint a picture that’s far more nuanced than the usual “AI will replace everything” or “AI is overhyped” narratives. The real story is that AI is reshaping how we think about talent, productivity, and problem-solving in software development.
We’re not heading toward a world where AI replaces engineers. We’re entering an era where AI becomes the ultimate skill amplifier, separating those who understand both technology and human needs from those who only memorize syntax.
The winners won’t be the ones who fear AI or the ones who blindly embrace it—they’ll be the engineers who understand exactly where humans excel and where machines can take over the grunt work. The question isn’t whether AI will change software development; it’s whether you’ll be ready to leverage that change.
Follow me on LinkedIn for more content like this!