Keeping up with AI productivity , vibe coding is obsolete ,
Keeping up with AI productivity, vibe coding is obsolete, handling conflict
Three quiet signals from the TLDR Product Management desk this week. One says AI has made engineers faster than their managers can keep up. Another says "vibe coding" is already a relic. The third says the actual scarce skill in 2026 is the one nobody puts on a job description: handling conflict without torching the team.
What You Need to Know: HBR's new piece argues that AI has made engineering teams noticeably faster, but the bottleneck has shifted upward to managers who are still running reviews, OKRs and one-on-ones at human pace. Separately, Jeff Gothelf's read of Andrej Karpathy argues that "vibe coding is obsolete" because once the model handles execution, the value migrates to problem definition — which is product management by another name. And Bakadesuya's guide on handling conflict is a reminder that the hardest part of any of this is still the people in the room.
Why It Matters
- If your review cadence hasn't changed in 18 months, you are now the slowest part of the system you manage.
- "Vibe coding" (describe what you want, accept what comes back) is being replaced by agentic engineering — and the role that used to be "coder" is collapsing into "product manager who can prompt."
- The 2026 competitive moat is taste, judgment and the ability to argue well in a room. That is not a tooling problem.
- Conflict handling is not soft skills. It is the throughput cap on every team you are standing up.
- Stop measuring output. Start measuring whether the work creates customer value, because the model can fake the first one.
What Actually Happened
Managers Are Now the Bottleneck
HBR's "Managers Are Struggling to Keep Up with the AI Productivity Boom" is the clearest framing I have seen. The argument is simple: the tools have made individuals 2–5x faster on a tight scope, but the coordination layer above them — planning, prioritization, career conversations, even weekly syncs — still runs on the same calendar. So the team ships faster, the manager drowns, and the org quietly regresses to the manager's pace.
The piece lands three prescriptions. Set clearer direction (because "go faster" is not a direction). Focus attention on the work that matters (because AI gives you more work, not better work). And build faster feedback loops without micromanaging (because status updates are now a model problem, not a people problem). None of this is novel. What is novel is that the cost of not doing it just doubled. (HBR)
Karpathy Says Vibe Coding Is Obsolete. He Is Right.
Jeff Gothelf's "Karpathy said vibe coding is obsolete — what he described instead is product management" is the cleanest restatement of the moment we are in. Karpathy's point, as Gothelf reads it, is that describing a feature and accepting whatever code comes back was a 2024 trick. In 2026, the model can build the whole thing, so the work is no longer "what should this do" — it is "what should we build, for whom, and how will we know it worked."
That is a product manager's job description. Gothelf is blunt: the "vibe coding engineer" was a transitional role, and the people who survive the transition are the ones who can frame a problem tightly enough that an agent can attack it. (jeffgothelf.com)
Handling Conflict Is the Skill of the Year
Bakadesuya's "How To Handle Conflict" lands on the same week for a reason. The argument is that conflict is only useful when it serves a clear goal, and most of what passes for conflict in a tech org is goal-less, ego-driven and a tax on shipping. The fix is procedural: focus on the real issue (not the loudest person), anticipate the other side's response (because surprises make fights worse), and stop before the fight starts controlling the room.
The 2026 reading: as agentic work makes individual output cheap, the only thing that compounds is the team's ability to disagree productively. (bakadesuyo.com)
AI Workspaces Eat SaaS Tools
Jason Calacanis's thread "From SaaS Tools to AI Workspaces" is the structural version of the Karpathy take. If agents become the unit of work, then the SaaS tool you log into is the wrong layer. The winners in 2026 will be the people who can describe what to build, manage agents, and turn generic model output into something specific to a customer. (threadreaderapp.com)
The Take
Look, I have been a manager of managers. The HBR piece hurt to read because it is right. The single highest-leverage thing I did in 2025 was delete recurring meetings and replace them with written direction — and the team got 30% faster in a quarter. AI did not do that. The discipline of writing clear direction did.
On Karpathy: the people who learned to "vibe code" in 2024 were the right people for 2024. In 2026, the right people are the ones who can sit in a customer call, hear a fuzzy problem, and ship an agent brief by Friday that produces a working prototype on Monday. If that is not you yet, that is the gap to close. Coding is no longer the moat. Framing is.
The conflict piece is the one I am going to send to every manager I know. Because the org that handles disagreement well is the one that does not need a manager to mediate every cross-team decision. Everything else in this digest is downstream of that.
Quick Summary
AI productivity is real, but it is making managers — not engineers — the constraint. Karpathy says "vibe coding" is dead; what replaces it is product management. And in 2026, the most underrated engineering skill is the one nobody hires for: handling conflict.