You Can't Prompt the Room: The Last Skill AI Won't Replace - Balázs Horváth, VisualLabs
TL;DR
17 of 21 agent ideas failed: VisualLabs' internal hackathon revealed that most AI projects die not from technical failure but from building something nobody needs.
Code is cheap, requirements are expensive: The bottleneck has shifted from writing code to accessing stakeholders and eliciting what actually matters.
AI gives you average answers by design: Without proper guidance, AI tends to replicate what already exists rather than innovate beyond the "faster horse" problem.
User stories still matter because AI was trained on them: Structured formats like persona-need-why give AI recognizable patterns, producing better outputs than vague prompts.
Shift your smartest people upstream: Before AI, your best people wrote code. Now they should be closest to customers and business problems.
The Breakdown
When VisualLabs ran an internal hackathon with 21 agent ideas, 17 were abandoned because they created no business value. Only four survived to reshape how the company works today. The lesson: writing code is no longer the bottleneck in software development, and figuring out what to build is now the expensive part.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.
Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.