Launch HN: Bloomy (YC S26) – AI-powered mastery learning for K-12

Hello Hacker News! We are thrilled to introduce Bloomy, a venture backed by YC S26 that is reimagining the fundamental architecture of K-12 education.

The Problem: The "Factory Model" Failure

For decades, schooling has operated on a fixed-time, variable-learning schedule. In a traditional classroom, the teacher moves to Chapter 2 on Tuesday, regardless of whether 30% of the class failed to grasp Chapter 1. This creates "knowledge gaps" that compound over time, eventually making success in advanced subjects impossible.

Education is a conveyor belt. Education should be a ladder.

Our Solution: AI-Driven Mastery Learning

Bloomy implements Mastery Learning, a pedagogical approach where students must demonstrate a deep understanding of a concept before progressing. While this was previously impossible to scale in a 30-student classroom, our AI engine makes it feasible.

How Bloomy Works

Our platform doesn't just provide answers; it acts as a Socratic tutor. It identifies the exact point of friction in a student's mental model and pivots the instruction in real-time.

"The goal isn't to finish the curriculum by June; the goal is for every single child to actually understand the curriculum by June." — The Bloomy Mission Statement

The Mastery Loop

Below is the logic flow our AI follows to ensure no student is left behind:

Technical Deep Dive

Bloomy utilizes a proprietary combination of Large Language Models (LLMs) and structured Knowledge Graphs. We don't rely on the LLM to "know" the curriculum; instead, the LLM navigates a mathematically verified map of prerequisites.

The Mastery Formula

We quantify mastery using a probabilistic model. A student is considered to have mastered a concept when the probability of a correct response $P(C)$ across varying difficulty levels exceeds a specific threshold:

$P(M) = \prod_{i=1}^{n} (p_i)^{w_i} > \tau$

Where:

• $p_i$ is the probability of success on task $i$.
• $w_i$ is the weight of the task's complexity.
• $\tau$ is the mastery threshold (typically $0.90$).

Data Structure Example

Internally, a student's progress is tracked via a state object similar to this:

{
  "student_id": "user_8821",
  "current_module": "algebra_1_linear_equations",
  "mastery_score": 0.84,
  "bottlenecks": ["negative_coefficients", "distributive_property"],
  "next_recommended_action": "remediate_distributive_property"
}

Traditional vs. Bloomy Learning

Implementation Roadmap

We are currently rolling out to beta schools. Our onboarding process for districts looks like this:

• Curriculum mapping to state standards
• Teacher dashboard integration
• Full-scale student deployment
• Longitudinal efficacy study

Getting Started

To integrate Bloomy into your classroom, you can initialize the client using our SDK:

npm install @bloomy/core

Then, set your learning_rate and strictness_level in the config: const session = bloomy.init({ strictness: 'high', mode: 'socratic' });

We would love to hear your thoughts on the future of personalized education. Are we moving toward a world where the "grade level" becomes obsolete?

We'll be hanging out in the comments to answer your questions!