Agent Experience Hackathon: Strands Decider
October 16, 2026 · 12:00 AM PDT – October 23, 2026 · 11:59 PM PDT
Online
Public
4375 participants
About
Build a working agent on AWS that does something cool, using Strands Decider 2B to make a meaningful decision, for a chance to win $5,000 in AWS credits and an AWS Builder swag bundle.
Strands Decider 2B is a small, open, experimental decision model from AWS. It does not generate text. It reads a state and a typed question and returns a choice, a yes or no, or a score, each with a calibrated confidence number, in about 115ms on hardware you already own. Pair it with Amazon Bedrock for the generative part of your agent, hand Decider 2B a decision, and show us what it lets your agent do. You pick the problem and the audience. Terms Apply.
Who can enter: Open to builders 18 or older with a Builder Center profile. See Rules for country and employee exclusions.
How to qualify: Every submission must include:
- An agent that uses Strands Decider 2B for at least one meaningful decision, paired with Amazon Bedrock
- A written article on Builder Center describing the agent, what it does, and the decision you handed to Decider 2B
- A link to a public repository with the agent's source code and at least one reusable Decider 2B decision pattern others can use in their own agents
- A short, publicly available demo video, under three minutes, showing the agent and its Decider 2B decision working
- One decision pattern (below) that your submission demonstrates
- An original agent that has not been published before
Build with an experimental model: Decider 2B is an innovation and experimentation starter, not a production product. Build something cool first. If you also measure a latency or cost improvement from moving a decision off the LLM, that is welcome, but not required.
How submissions are judged: Judging runs in two rounds:
- Gate 1, AI and human scoring: every qualifying submission is scored on the rubric below. The top 100 submissions advance.
- Gate 2, human judging: a panel of AWS experts reviews the shortlist and selects three winners.
The rubric is weighted toward creative experimentation:
- Creativity and cool factor (30%): How interesting, delightful, or clever your agent is.
- Effective use of Decider 2B (30%): How well your agent uses Decider 2B with Amazon Bedrock, including the quality of your reusable decision pattern. A measured latency or cost improvement counts as a bonus here.
- Communication quality (25%): How clearly your article and video explain the agent, the decision, and what you learned.
- Originality (15%): The originality of your agent and the decision you chose.
Decision patterns: Every project hands Decider 2B at least one meaningful decision, made once or made over and over. Repeated, high-frequency decisions are where the speed and cost benefits show up most, but a single meaningful decision is just as welcome. Declare which pattern you demonstrate:
- Tool selection: The agent decides which of several tools to call from the user's input.
- Guardrails: Before a tool call, verify the arguments are grounded in what the user actually said.
- Policy: Classify an incoming request into a category before the agent acts on it.
- Model routing: Decide whether a question needs a frontier model or a smaller one.
- Context and memory: Score which memory or context chunks are relevant before injecting them.
The reusable asset: Every submission includes at least one reusable Decider 2B decision pattern another builder can drop into their own agent: the typed question and option set, the wiring that puts it in the agent's path (for example a Strands InterventionHandler), and a short note on how to use it. Together these become a shared library of decision recipes the community can reuse.
How to tag your submission: Every submission carries two tags so it is routed for judging and grouped by decision pattern: 1) the required #strands-decider-2b tag and 2) one decision pattern tag identifying the pattern demonstrated (#tool-selection, #guardrails, #policy, #model-routing, or #context-memory). Add both before the deadline.
Prizes: Three winning projects will be selected. Each receives $5,000 in AWS credits and one AWS Builder swag bundle worth about $600 (jacket, backpack, keyboard). Each winning participant also receives an official winner's certificate. Total prize pool: $16,800.
NO PURCHASE NECESSARY. VOID WHERE PROHIBITED.
| Hackathon Milestone | Date |
|---|---|
| Launch | October 16 |
| Projects due by | October 23 |
| Gate 1 (AI and human scoring) | Week of October 26 |
| Gate 2 (human judging) | Week of November 2 |
| Winners Announced | Week of November 9 |
Awards
1) 1 winner receives $5,000 in AWS credits and ~$600 exclusive swag bundle
$5,600
2) 1 winner receives $5,000 in AWS credits and ~$600 exclusive swag bundle
$5,600
3) 1 winner receives $5,000 in AWS credits and ~$600 exclusive swag bundle
$5,600
