June 28, 2026
Petals, panic, and a title fight
Build Yourself Flowers
A soul-searching AI talk sparked a surprise fight over the title, not the tech
TLDR: Vicki Boykis used a personal conference talk to ask whether careful machine-learning work still matters in the age of AI shortcuts. Commenters, however, got stuck on the poetic title, splitting into camps over whether it was thoughtful and fitting or just annoyingly vague.
Vicki Boykis brought a big, thoughtful mood to her keynote, “Build yourself flowers”, tracing her path through years of building machine-learning systems and asking the question haunting a lot of people in tech right now: if everyone is rushing to use chatbots and auto-generated code, does careful craft still matter? She even broke down how the talk itself was made — handwritten draft, local transcription, AI cleanup, then old-fashioned human editing — in a very human reminder that the final polish still came from, well, a person.
But in the comments, the crowd immediately grabbed the steering wheel and swerved into a completely different lane: the title drama. One reader was openly exasperated, basically saying, “Why are Hacker News titles always so vague and artsy?” and threatening the digital equivalent of flipping the table by flagging it. For them, this wasn’t poetic, it was communication malpractice. The hottest take? The article should have had a plain-English title like a workshop flyer, not a metaphor that made readers guess what they were clicking.
Then came the defense squad. Another commenter jumped in with context, explaining that this was an edited conference talk and more of a meditation on the future of the field than a how-to guide. Translation: relax, it’s reflective on purpose. So yes, Boykis was talking about the identity crisis of modern AI work — but the community turned it into a mini culture war over whether the internet has forgotten how to name things clearly. Classic comments section behavior.
Key Points
- •The article is an edited transcript of Vicki Boykis’s keynote at the Applied Machine Learning Conference in Charlottesville, Virginia, in April 2026.
- •Boykis describes a production workflow that used handwritten drafting, MacWhisper, local Whisper transcription, Gemini Flash 2.5 in Pi, and manual editing to turn the talk into a blog post.
- •She identifies herself as someone who builds machine learning systems and summarizes prior work at Tumblr, Automattic, Duo, Mozilla.ai, and Malachyte.
- •The talk centers on whether traditional machine learning and machine learning engineering still have a clear role amid growing use of LLMs and generative AI.
- •The excerpt also revisits her earlier context-window keynote and explains context windows, token limits, and the role of attention in Transformer models.