November 4, 2025
Blank JSON, full drama
Launch HN: Plexe (YC X25) – Build production-grade ML models from prompts
Push-button AI? HN asks: Where are the inputs, the cleaning, and the code
TLDR: Plexe says it can build production-ready AI models from plain-English prompts. Commenters loved the idea but pounced on unclear inputs (that blank JSON), missing code in exports, and data-cleaning gaps—asking if it can even admit when simple averages beat AI—making trust and transparency the real battleground.
Plexe hit Launch HN promising a magic button for AI—tell it what you want, get a production‑ready model with clear metrics and explanations. The crowd was intrigued, but the vibe quickly turned into “cool demo, now show receipts.” One early tester cheered the idea then got stuck on a mystery moment: the sample request showed a blank JSON for inputs, sparking jokes about a “guess-the-fields” API while their training job lingered in “processing.”
Questions piled up fast. Fans wanted simple answers: Can it handle spreadsheets, photos, text, and audio? Will it actually clean and label messy data, or is that still on the user? Can the agent be brutally honest and say, “you don’t need AI—just use an average”? And yes, someone asked for an explainer agent for model decisions, not just shiny charts. Another user said the tool gave useful guidance and code—then the export left the code out entirely, like ordering pizza and getting only the box. Meanwhile, a seasoned voice shrugged: the real work is still data wrangling, forever.
So the plot: Plexe’s promise of transparent, ready-to-deploy AI meets a community demanding concrete inputs, end‑to‑end data handling, and exports that include everything. Hype is high, patience is low, and trust will hinge on whether Plexe can turn that blank JSON into real, working answers. Read the thread on HN and check Plexe at their site
Key Points
- •Plexe enables users to describe desired model behavior and data in plain language and builds production-ready ML models.
- •The platform provides clear performance metrics, training details, and explanations to foster trust in predictions.
- •Users are encouraged to be specific about prediction targets and data sources when configuring models.
- •Plexe is positioned as an end-to-end solution from first model creation through deployment at scale.
- •The goal is to turn organizational data into engineered AI solutions tailored to exact business challenges.