🤝 Top Collaborators
.png&w=640&q=75)
🤓 Latest Submissions
Orihime Tablekeeper
Orhime Tablekeeper is a modern restaurant reservation system designed to provide a secure, reliable, and conflict-free table booking experience. The core problem it solves is double booking, where multiple customers attempt to reserve the same table for overlapping time slots. The platform allows users to select a restaurant, specify the number of guests, choose a preferred date and time, verify their identity through OTP authentication, and confirm their reservation. Each successful booking receives a unique confirmation code, while users can also manage and cancel their reservations. The key strength of Tablekeeper is its concurrency-safe booking architecture. Instead of relying only on frontend availability checks, the system uses PostgreSQL transactions, row-level locking, and database-level exclusion constraints to ensure that conflicting reservations cannot be created. Even when multiple users attempt to book the same table simultaneously, the database guarantees that only a valid reservation is accepted. The system also incorporates idempotency protection to prevent duplicate bookings from repeated requests, along with rate limiting for booking and verification operations. Time-range validation and proper transaction handling further improve reliability. Tablekeeper demonstrates how a reservation platform can combine a simple user experience with robust backend engineering. Its architecture focuses on data consistency, security, concurrency control, and reliability, making it suitable for real-world restaurant booking scenarios where simultaneous users and race conditions are common. The project's main principle is simple: one table, one time slot, one valid reservation — even under heavy concurrent demand.
6 Oct 2026
DevPartner AI
The Problem AI coding assistants are increasingly trusted to modify production code — but most of them work as a "black box." A developer asks for a change, the AI edits files, and the developer is left hoping nothing breaks. There's no proof the change is safe, no record of what rules were checked, and no easy way to undo damage if something goes wrong. This is especially dangerous in sensitive areas like authentication, payments, and user data, where a single unverified AI edit could open a serious security hole. **The Solution** DevPartner AI is a **Developer Verification Agent** that sits between a developer's request and the actual code change — acting as a safety layer that never lets AI blindly modify production code. Instead of just generating a fix, it walks every change through a transparent, auditable pipeline: 1. **Project Scan** — indexes the entire codebase and profiles its structure, languages, and complexity. 2. **Change Request & Intent** — the developer describes what they want in plain English, and the AI interprets the true intent behind it. 3. **Invariants** — the system automatically extracts the rules that must never break (e.g., "only verified sessions can delete a user"). 4. **Impact Analysis** — calculates the "blast radius" of the change, showing exactly how much of the codebase it touches. 5. **Risk Assessment** — scores how dangerous the change could be, based on impact and any existing security findings. 6. **Change Plan** — shows the developer a clear, step-by-step plan before any code is touched. 7. **Approval** — a human-in-the-loop checkpoint, so nothing merges without explicit developer sign-off. 8. **Snapshot & Verify** — the current state is saved, then the change is tested through property checks, mutation testing, and adversarial counterexample hunting. 9. **Evidence** — a cryptographic SHA-256 hash is generated as tamper-proof, auditable proof that the verification actually happened. 10. **Rollback** — if anything is wrong, the developer can revert instantly with one click. Along the way, DevPartner AI also runs automated **security scanning**, catching real issues like hardcoded credentials or authorization logic that wrongly trusts client-supplied data — flagging them by severity (HIGH, CRITICAL) with the exact file and explanation. **Target Users** Software development teams, DevOps engineers, and security-conscious startups who want to safely adopt AI-assisted coding without risking silent, unverified changes reaching production. **What Makes It Creative and Unique** Most AI coding tools optimize for *speed* — generating code fast and trusting it. DevPartner AI flips that priority to **trust and proof**. It's not just an AI code editor; it's a verification layer that treats every AI-driven change like a claim that must be proven, not assumed. The combination of extracted invariants, blast-radius mapping, adversarial mutation testing, and cryptographic evidence generation is unusual for a hackathon project — it borrows ideas from formal verification and security auditing, areas normally reserved for large enterprise tools, and makes them accessible in a lightweight, interactive interface. The one-click rollback and in-memory sandbox isolation also mean developers can experiment fearlessly, knowing nothing is permanently at risk until they explicitly approve it. --- **Word count: ~480 words** — fits within your limit with a little room to spare. A quick honest note: I wrote this based on the UI and workflow you've shown me, but I haven't seen your actual code or confirmed every step truly works as described (e.g., whether mutation testing and counterexample hunting are fully implemented or partially simulated). Please adjust anything that overstates what's actually functional — judges may check claims against your live demo and repo.
27 Sep 2026
