Quick note: If this is the first email you’ve gotten from me in a while, that’s because my newsletter provider had a bug where my emails weren’t being sent properly (for literally six months). 😭
But we’re back! Each week, I share a combination of technical content for AI/ML roles (roadmaps, learnings, recommendations — that kind of thing) and mindset tips for success in challenging fields.
If that’s not your thing, there’s a link to unsubscribe at the bottom. No hard feelings. But I hope you’ll stick around!
In This Newsletter
Thank you for your support <3
Facts. Without Hyperbole. In One Daily Tech Briefing
Get the AI & tech news that actually matters and stay ahead of updates with one clear, five-minute newsletter.
Forward Future is read by builders, operators, and leaders from NVIDIA, Microsoft, and Salesforce who want signal over noise and context over headlines.
And you get it all for free, every day.
What 500 Top Engineers Taught Me About the Future of AI in Software Development
I just came back from the Pragmatic Summit, a meeting of five-hundred top 1% engineers hosted by Gergely Orosz, author of The Pragmatic Engineer (the world’s most-read engineering newsletter).
Here's what I learned.
AI adoption is nearly universal—but results vary wildly
Brand new data shows that roughly 93% of developers are now using AI tools. Among those who use them, the average time saved is four hours per week. AI-authored code has climbed sharply quarter-over-quarter: 22% in Q3 2025, jumping to 27% between November 2025 and February 2026. Teams are also reporting that AI cuts onboarding time in half.
But there's no typical experience. AI is a multiplier of whatever is already happening on a team. Some companies have seen customer-facing incidents double while others have cut them by half. Dysfunctional teams are now dysfunctional faster, and strong teams are strong faster.
Fundamentals matter more than ever
Every person I spoke with emphasized the ongoing importance of fundamentals. If everyone has access to the same AI coding tools, those who succeed will be the ones who can prompt effectively. And to prompt effectively, you need to understand architecture, design trade-offs, and lower-level abstractions.
So yes, you should still learn to code.
One analogy I heard: the best mechanical engineers have been machinists. Knowing what's happening at a practical level helps you design better systems.
Managers also expressed frustration at the lack of fundamentals among junior engineers. All engineers need to understand a few abstraction layers down. Architecture knowledge helps you write good prompts and use AI effectively, like how to work in parallel with multiple agents Same idea as strong teams vs. weak teams: AI just makes you faster at whatever you were doing before, good or bad.
Hiring is changing fast
I sat in a roundtable with 20+ hiring managers from startups and companies like Ripple, Coinbase, and Capital One. Here's what they shared:
Moving away from LeetCode: Companies are shifting toward simulations of actual work rather than algorithmic puzzles.
Frustration with the "AI arms race": Hiring managers described frustration with AI-generated resumes submitted by bots, then screened by AI on the company side. Some are introducing filters like requiring resumes via a POST request, or leaning harder on referrals and networking, but they know they’re missing good candidates. Cheating is rampant—one manager mentioned hiring someone from remote interviews only to have a different person show up to work. Some companies now ask for Loom videos to prove candidates are real.
Testing AI-native skills: Many are turning to take-home assignments where candidates are explicitly asked to use AI. They follow up with live pair programming where you extend the project. They're evaluating prompting ability, spec-driven development, code review, testing, debugging, and the ability to correct the LLM. Several hiring managers said they're literally looking for whether a candidate has the patience to read everything. They want people who are not outsourcing their thinking to AI.
Going deeper on past work: Others are asking detailed questions about GitHub projects. This is becoming more common—managers want to see real evidence of your work.
Brownfield testing: Some companies create fake repos with deliberate bugs and ask candidates to find and explain them using AI, or to pair-program a code review.
Hunting in open source: One manager said he looks at commits on open-source repos and headhunts from there.
They're looking for "T-shaped" engineers. It's no longer considered reasonable to "only" do frontend, for example—everyone needs to be full-stack with some specialization since the tools have made general engineering skills accessible, at least at a surface level.
Valuing product thinking: They want engineers who can navigate probabilistic, chaotic environments with strong product intuition.
The real bottleneck isn't technology—it's organizational change
It doesn't matter if dev teams want to use agents if there aren't appropriate change management systems in place. Entire organizations need to be revamped and engrained ways of working need to change. This can't happen without strong leadership and organizational buy-in.
To see real impact from agents and AI, teams must have clear goals and measurements. Change management will be key. You won't see economic gains without huge changes—and those changes need to be done well.
No one knows what the future looks like
I asked Chip Huyen, author of AI Engineering, what her advice would be for students considering AI engineering careers over the next two to three years. Should they bother learning given the pace of change?
She said engineering has never been about learning to code—it's about solving problems, and problems will continue to exist. But she also told a story about a student choosing physical therapy over CS, so I don't think she's completely sold on the long-term outlook either…
Relatedly, Tibo Sottiaux, CTO of Codex, was asked what the software engineering job will look like in two years. His answer was that two years is too long for any predictions. He had some guesses for six months out, but that was it.
If these folks don’t know what to expect, I definitely don’t.
But they don't expect mass layoffs (…they say…)
Multiple leaders expressed that they're viewing efficiency gains with excitement about all the new things their teams can build—not as an opportunity to lay everyone off. Maybe they were being politically correct for the crowd.
But in my own experience, my workload is much higher now. I'm able to deliver more value more quickly, which makes me more valuable to my organization, not less. I've also needed to upskill coworkers to take over some of my work so we can do more. I am not seeing interest from management in doing less of this work given that we're making more money for our companies.
Roles are blending
Vibe coding has allowed PMs and designers to build things themselves—the Codex team said that non-tech people at OpenAI are now coding as much as tech people were six months ago. Meanwhile, less time actually writing code means more time in product design and evaluation, so engineers need stronger product sense.
I've noticed this personally. I'm spending much more time coordinating, thinking about product decisions, legal considerations, and cross-team alignment, and less time coding. Managers are able to contribute more technically too.
Side note: Some folks think fewer managers will be needed in the future once we have more AI coordination. Something to consider if you're an IC thinking about moving into management…
The biggest gains are still ahead
Right now, agents mostly help with writing code — but engineers still spend the majority of their time in meetings. The real transformation will come from fixing system-level problems and freeing up developers to actually develop.
And the ceiling is higher than most people realize. Top teams like Codex are already running swarms of agents in parallel overnight. They argue that if you're getting mediocre results from AI coding assistants, the bottleneck is probably your prompting skills, not the models themselves. We're still early, and there’s a lot more to come.
Feelings are high and mixed
A lot of people are scared about the future of their jobs. But I also heard the phrases "coding is fun again" and people expressing genuine joy more than you'd imagine. There's excitement and optimism, alongside a healthy dose of skepticism.
The Best AI Engineering Books for Beginners
A few weeks ago I posted a video sharing the 10 AI/ML books I’m reading in 2026, and I got MANY requests in the comments for a video on books that are better for beginners. So here is that list!
I focused only on practical books that are fun, engaging, and will get you from beginner to professional in the shortest possible time.
Links to all the books are in the video description, or you can grab them from the blog here!
Book 1:1 Coaching Before Prices Increase
Like I shared in last week’s newsletter, I’m launching a learning community in a couple of months!
This will be a place where you can ask questions and get feedback from me in a group setting for an affordable rate. I’m still working out the exact details, but it’ll definitely include a-sync and live support, learning resources, accountability, and networking.
I’m sharing this because I’m excited, but also to give you a heads up: Once the community launches, my rates for one-on-one coaching will go up significantly. So, if you want to meet to discuss your career individually, I’d highly recommend booking now!
Thank you all for your support. I’m really looking forward to this and I hope it will be helpful. <3
Last thing — Appa says hi!
Want to chat 1:1? Book time with me here.
Forwarded this email? Sign up here.
Note: This email may contain affiliate links. If you make a purchase I may make a small commission, at no cost to you. Thank you for your support!