ContextPulse AI
Target Users: - Software Developers & Engineers (experiencing high cognitive disruption) - Engineering Team Managers (facing hidden operational blind spots) Core Problems Today: 1. The Context Recovery Tax: Developers waste 2–4 hours daily rebuilding mental focus and workspace context after routine breaks or task switches, destroying their flow state. 2. Surveillance Fatigue & Privacy Violation: Existing productivity tools act as intrusive spyware (logging keystrokes, capturing screens), creating toxic workplace distrust. 3. The Management Blind Spot: Engineering leaders remain completely blind to cross-team blockers until deadlines fail, resulting in misguided blame instead of systemic fixes.
Preliminary score
- Value for the user15.5 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI17.5 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing11.5 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility6.5 / 15
Data requirements, running costs and a clear next step.
- Originality6.0 / 10
A meaningful difference from common solutions.
What the judges wrote
- Claude Opus 5.5Preliminary54/100
- Strengths
- Your privacy design does real work in the code. The Pydantic schemas reject any event with a true privacy flag or extra fields like source_code with a 422 error, and the manager dashboard hides groups until K = 5 responses. Your Gemini reminder is checked in code for invented file, function, error or PR names and falls back to safe text, a concrete fix for a failure you actually saw. The system also runs end to end, with an IDE simulator, live SSE updates, AZ/EN support, pytest coverage, Docker and CI.
- To improve
- Your impact claims (2–4 hours lost daily, 25–35% productivity gain, '0-second' recovery) have no evidence behind them. Even a small trial would be far more convincing, for example a few developers rating AI reminders against their own recall. The AI's contribution is thin: the app works with fallback text and no key, so show side-by-side examples of Gemini reminders versus the fallback and how often validation rejects output. Add running costs per developer and a concrete next step, such as a VS Code extension or a GitHub webhook to replace the simulator.
- GPT-6 AstraPreliminary60/100
- Strengths
- Your project connects a concrete developer problem—resuming work after interruptions—with team-level blocker visibility rather than individual surveillance. Your documented prototype covers metadata ingestion, bilingual reminders, live dashboard updates, and AI output validation with a safe fallback. Your testing account identifies specific failures, including fabricated function names and overly strict schemas, rather than relying only on success claims.
- To improve
- Your next demonstration should show a complete metadata-to-Gemini-reminder workflow, distinguish live AI from fallback output, and reconcile the Gemini 2.0 versus 3.5 model settings. Publish test results and measure task-resumption time against manual notes or IDE history before claiming zero-second recovery or productivity gains. Your K=5 protection needs evidence of five distinct participants—not merely five responses—and tests for information leakage through repeated submissions and API/SSE access. Add estimated per-developer running costs, metadata retention rules, and a concrete next step toward a real IDE/PR integration pilot.
The team
Nuran EliyevLead
designer
Elgun Huseynli
developer
Mikayil Movsumzade
developer