@mitchellkosowski.bsky.social

@mitchellkosowski.bsky.social

7 Followers 11 Following

*opens Claude* “Tell me how to feel about this meme.”
Me 🤝 me
Yeah, I know ball.
No thoughts, just vibes 😎
Daily affirmation: AI isn't replacing me.
Fun team-building activity to try out!
Live laugh lowkey panicking 😌
Okay, but what's the conversion rate on this one... Asking for a friend.
But... I thought ChatGPT and I were friends? And that, your honor, is how the customer database ended up in the prompt.
13.8 billion years of industry experience for an entry-level role? OK, expectations are getting a little unreasonable...
Avoiding 4 hours of work with 10 hours of psychological warfare and de-bugging induced character development. It's simple math, people!
I prefer to call it open-source collaboration and “using your resources.” The resource just happens to be a Stack Overflow thread from 2014 written by someone named xXDarkCoderXx with 3 upvotes and 7 followers. 😌
I've never met a pot I didn't want to stir...
Most engineering orgs were designed around human-only workflows. Now agents are handling larger parts of development, testing, debugging, & deployment, but the org structures around them haven’t caught up.
Back in the olden days of early 2022...
Double check my math on this, but it looks like the enemy of my enemy is... my friend?
Proud to be part of a team that understands the weight of that responsibility and shows up with the precision it demands. Listen to my interview here: https://open.spotify.com/episode/577LsJdlRgnjWY6UVgam6x?si=ODpFMKBARsKLlqtMwTEKUw
Getting ghosted by users is my love language. <3
Research shows AI systems tend to agree with users far more than humans do, even when the user is clearly wrong. It feels supportive in the moment, but over time it weakens judgment and makes you question less. And that's a slippery slope. https://www.science.org/doi/10.1126/science.aec8352
Humans are so back!
Ah yes, the classic walking on clouds to processing them switcheroo.
The job market makes no sense. 67K+ engineering roles up ~78%, yet layoffs continue. Companies cut misaligned roles and hire senior talent. Everyone wants seniors, but fewer are being built. You can’t hire your way out.
Turn an idea into a clickable product in 15 mins with no code? “The prototype is the new PRD” sounds dramatic… until you try it. I gave it a go with Google Stitch. And it works. Now it’s hard to justify a 20-page PRD when you can just show the thing. 📩 https://tinyurl.com/bdh6r9ey
Yes, AI eliminates tasks but much of it now is repetitive work like boilerplate code. Meanwhile pay is going ⬆️ The role is shifting toward work that takes human judgment, like understanding problems, designing systems, making tradeoffs, and knowing what not to build. https://tinyurl.com/y5uzh3mu
And people say AI isn’t creating new jobs...
This one was a lil too #relatable. “Entry-level” now means mid-level output. Orgs want fully ramped engineers on day one. Just ones that someone else trained. So we get this loop: No one hires juniors ➡️ Everyone complains they can’t find talent At some point, someone has to hire for potential.
Nothing built character quite like spending three hours tracking down a segmentation fault that Python would have just... handled.
You can hire great engineers, then build a system where they need permission to do basic work. It happens when “autonomy” exists on paper, but not in the work. Redundant approval chains = No real ownership. Leaders have the autonomy to create autonomy. 📩https://substack.com/home/post/p-189553599
One developer panics over the complexity. Another installs a package and moves on. Neither is wrong. But only one is thinking about leverage. npm install didn’t kill frontend development. It removed a lot of unnecessary work and freed engineers to focus on what actually matters.
Companies say they want AI adoption. So they implement mandates and track logins. But opening an AI tool only proves one thing: Someone opened a tab. Real adoption is cultural. 📩 https://signalvsnoisetechinsights.substack.com/p/your-ai-mandate-wont-work-try-this
Engineering teams build systems for technical failure all day long. But when the failure is human most teams do nothing. That’s the gap between individual EQ & team EQ. Tension has signals too. Treat it that way. Conflict retros > letting it silently slow the team down. https://tinyurl.com/2vh34wcw
Next time I meet with VCs I might start using “fresh environment initialization” just to see who nods confidently...
Potentially hot take: Most engineering problems aren’t technical (IQ). They’re human (EQ). Think misalignment, unclear ownership, tension, assumptions no one surfaced. You aren’t born with fixed EQ. It’s a skill you build the same way you build good software. https://tinyurl.com/fhstt2dj
65 years since its inception, COBOL still runs core banking, payroll, & government systems. Most institutions don’t replace it. The foundation still holds, so they can build around it. Strong leaders modernize while maintaining what already works. Full take here: https://tinyurl.com/4z6289cd
Before you default to “we need to hire more engineers,” pause. Take a walk. Clear your head. Then come back and ask: Will this actually solve the problem, or does it just feel productive? The pattern usually goes something like this: The team feels stretched. Deadlines slip. Frustration rises.
It’s funny because it’s true. At some point, experience teaches you that “simple” is rarely simple up front. It’s usually the result of: Understanding the real constraints ...anticipating future change ...and choosing where not to be clever Abstraction isn’t the enemy. Premature abstraction is.
If your team needs constant checking, approvals, and reminders to ship, the problem isn't motivation. It's your management system. Most ICs struggle in their first EM role bc they keep using IC instincts. That usually means tickets over outcomes. Control over expectations. Activity over impact.
If your Git merges look like this, the problem is not Git. It’s your process. Most shipping pain follows the same slow descent into madness: Work sits too long ➡️ PRs grow too big ➡️ Context disappears ➡️ Merges get risky ➡️ People hesitate
Future You will be asked why they programmed it that way. Be kind to them.
AI is an implementation detail not a value prop. At the end of the day, customers buy outcomes not the models that drive them. If you can't sell the pen without saying “AI,” you probably don't understand the problem well enough yet. Defining what you're actually trying to get done is step #1.