Alex Jauch (@ajauch) header
Alex Jauch (@ajauch) profile picture

I've spent 30 years building enterprise software. Now I teach builders how to do that with AI. https://teapawt.org/t.co/mM5gyB69Jg

Joined May 2009
Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)Photo by Alex Jauch (@ajauch)
The verdict is in: Measuring severity-weighted issues per 1,000 Lines of Code, CI/CD cuts vibe-coded repo issues in half (1.50 → 0.74) but has almost no effect on indie repos (1.20 → 1.35). AI-generated code benefits dramatically from automated quality gates. Without CI/CD, vibe-coded repos are the worst box; with it, they're the best.
1
89
Next time you hear someone dismiss a software product as an “LLM wrapper” just remember that Nvidia doesn’t make chips, Nike doesn’t make shoes or clothing and Coke doesn’t bottle soda. Building a business isn’t about owning the value chain end to end. It’s about building something your customers value.
19
“People don’t want to buy a quarter-inch drill, they want a quarter-inch hole.” --Theodore Levitt This was true in 1960 and it's true today. Your customers don't want agents. They want a business problem to just go away. Solve the problem, don't sell the tool.
25
For the first few years, cloud was mostly hosting with a better UI. The real change wasn't the servers. It was that a developer could get a database without filing a ticket. Ask yourself what the equivalent unlock is with AI. It probably isn't the part being demoed.
19
AI is the new outsourcing. Same pitch, same spreadsheet, same executive enthusiasm. Twenty years ago it was labor arbitrage. The firms that treated it as a cost story got what they paid for. The ones that treated it as a capability story did fine. The industry obsessed with innovation runs the same movie every decade.
23
Every company says it wants a culture of experimentation. Then they go and promote people who ship things on time and under budget despite the odds. Here's the test: when did you last promote someone for killing a project? If never, you don't have an experimentation culture. You have a hero culture with a nicer name.
23
Before your company adopts any technology, answer these: What problem are we solving? Who owns the outcome? How will we know it worked? What does it cost at scale? What happens if we're wrong? Who has to change how they work? The last one is where almost everything dies.
1
31
An MIT report on AI in business found the overwhelming majority of enterprise GenAI projects produced no measurable financial return. I was not shocked. I watched the same thing happen with private cloud, with big data, and with digital transformation. The technology was never the problem.
23
What happens when you actually measure the quality of AI coding agents? alexjauchconsulting.com/blog…
1
25
I ask every PM who joins my org what their career goals are. A shocking number tell me no manager has ever asked them that. Ever. Senior PMs, years in, never once. If you're a PM: has any manager asked you?
41
My goal is to be the dumbest guy in the room. In my 20s I worked hard to be the smartest. It worked. My reward for solving impossible problems was more impossible problems, and I'd convinced everyone I was the only one who could do them. I was indispensable and completely burned out. So I started investing in the people around me instead. Workload down. Output up. Quality up. I'm no longer the smartest person on my team and I love it.
41
"Build feature X, I think it's a good feature" is not a request I can act on. "We need to cut churn 50%" is. I may have no idea how yet, but that's a fair thing to ask me, and I can go find out. Bring outcomes to PM, not opinions.
28
Every PM team I've led has had a public stack rank. I'm maniacal about it. Not because I can please everyone — precisely because I can't. Anyone in the company can see what we're doing and why. They can disagree. They can't be surprised. That alone kills half the noise.
1
52
Years ago I wrote a course for my sales team called "How to Operate Your Friendly Neighborhood Product Manager." Deliberately a little silly, because the point wasn't to scold them. They wanted something from PM, and I was going to tell them exactly how to get it. If your org keeps interacting with you badly, that's a training problem, and it's yours to fix.
33
I have the receipts: 1,800 GitHub repos later and I can tell you with high confidence that "AI Slop" is just an excuse for crappy development practice: AI will produce worse results. If you don't bother with testing, planning, architecture or any best practices. With them? The results are better than writing by hand. medium.com/@ajauch/does-vibe…
30
Free advice for vibe coders you didn't ask for: Ignore the time estimates AI gives you. They're wildly off, exactly like a junior's. Nothing wrong with that, just don't build a plan on them.
27
Free advice for vibe coders you didn't ask for: Download the official best-practice doc for whatever you're using. Drop the PDF in your docs folder. Tell the AI to compare your implementation against it. You'll get a concrete punch list of what's wrong with your implantation for free. Take that punch list and stack rank it. P0s now, P1s and P2s in the backlog. Not everything the audit finds is urgent, and treating it all as urgent is how you go insane.
21
A conversation I have had, more than once: "We know customers want X." "How do we know?" "They told us." "Why do they want it?" "It doesn't matter, we need to do it." "Can you point me at one customer so I can ask?" "I'll get back to you." Spoiler alert: They did not get back to me.
1
114
The PM is not a feature vending machine where you insert requirements and receive features. Your job is to own the customer outcome. That means that you say "no" more than you say "yes" to internal stakeholders. You only know you've succeeded when customers tell you by using the thing and paying for it.
32
Signs your team has a solution looking for a problem: Engineering builds prototypes nobody asked for. Leadership "vision" has drifted from customer reality. New customer problems get quashed automatically. PMs aren't meeting customers directly. I've been guilty of the third one. If you do it, at least know you're doing it.
29
If you're a PM, don't fall for the trap of committing to dates. Only engineering can ship a feature, so let them answer "when." What you owe leadership is a clear stack rank, an honest read on complexity, and a very rough sense of time. If X is six weeks out and lands in eight, nobody cares. If you called Y the most important thing on the roadmap and didn't ship it, you have a real problem.
24