Denis Murithi@deniskelvin92239

A customer places an order. The server records it, but the confirmation disappears on the way back. The customer sees “failed,” tries again, and creates a duplicate. Most testing tools can simulate a network problem. ReceiptGap makes the missing human consequence visible: what did the person believe, what did the system actually do, and what happened next? ReceiptGap is an interactive transaction truth lab for developers. Builders select a failure boundary, then compare a broken implementation with a repaired one. The app shows the customer-facing message, the server ledger, a step-by-step event trace, and a verdict. In the broken flow, a lost response produces a false failure and a second order. In the fixed flow, the client treats the outcome as unknown, checks the original operation ID, and shows the existing receipt instead of creating another order. The event trace can be exported as JSON, and the deterministic engine is covered by focused tests in the public repository. ReceiptGap is aimed at teams building checkout, bookings, applications, and other high-stakes workflows. The prototype uses synthetic orders and makes no real payments. The planned hackathon build will extend the simulator into a developer workflow that imports a synthetic trace, validates invariants, and generates a replayable regression fixture.

Delivery teams lose time when a failed drop-off becomes a vague chat message: “customer not there,” with no clear record of what the rider tried or what dispatch should do next. RouteProof gives riders a hands-free way to report the exception while they are still at the stop. An AssemblyAI Voice Agent asks short follow-up questions about the issue, contact attempt, customer response, and suggested next step. Its tool call creates a structured draft, not an automatic decision. The interface shows the live conversation beside the draft so dispatch can catch mistakes before approving it. The app also checks whether a short evidence quote appears in the rider’s transcript and flags any quote it cannot match. A missing-facts gate stops an incomplete draft from being approved. The prototype uses sample Nairobi routes and saves reviewed reports in the browser. It is designed to show how voice can make field reporting faster without letting a speech recognition guess silently change a delivery record.

Small businesses need short-form video content constantly, but most don't have the time, tools, or budget to produce it. Every TikTok, Instagram Reel, or LinkedIn post starts from a blank canvas. AdCollage removes that barrier by turning a business's own website into the source material for a finished promo video. Give the agent a URL and a target platform, and it scrapes the site for brand assets: logo, product images, color palette, and key copy—then selects a pacing and visual template suited to that platform's format (fast-cut vertical for TikTok and cleaner and slower for LinkedIn). It assembles those assets into a rendered video collage: synced transitions, on-brand color treatment, and text overlays pulled directly from the site's own messaging, so the output already sounds like the business, not a generic template. The rendering pipeline is built on Remotion, driven by an LLM layer that handles asset selection, pacing decisions, and copy extraction—reflecting a production workflow already used to generate short-form brand video for multiple live clients. The AMD Developer Cloud and ROCm stack powers the generation layer, with the full pipeline containerized for submission. Unlike static video templates, AdCollage is generative end-to-end: no manual editing, no stock template library. Every output is built directly from the target business's real assets and voice, which means it scales across any industry, market, or brand without pre-built templates.

Six Matches. One leaderboard. Build on AMD, earn points and compete for prizes.

🌎 Online Hackathon | 💻 Online Build: Sept 26 - Oct 5, 2026 A virtual, global hackathon. Build from anywhere in the world. Dark Factory is the official hackathon of WeAreDevelopers World Congress North America, presented by BAND, but you don't need to be in San Jose to take part. For one week, you build a software factory in BAND Desktop: a band of coding agents that plans work, implements it and checks its own results. 🎟️ Bonus for sign-ups: you are eligible for a free ticket to WeAreDevelopers World Congress North America in San Jose (Sep 23–25). If you do go, stop by the BAND booth: a surprise is waiting for you.

💻 Online Hackathon | ⏳ Build: 48 hours | 📅 September 25–27, 2026 The official IBM Bob 2.0 hackathon. Build what's next in AI-assisted development, alongside the developers and experts shaping it. ⭐ Get help from experts throughout the event 🤝 Join a team or build solo 💰 Prize pool: $12,000 🧑🏻‍💻 Register before the Kick-Off Stream to start building with Bob

🌎 Hybrid Hackathon | 💻 Online Build: September 10–16, 2026 | 📍 On-site: September 15–17, 2026 The official hackathon of AI Infra Summit — co-organized by Kisaco Research and lablab.ai. Build production-grade AI applications on the modern AI infrastructure stack, alongside 8,000+ attendees, researchers, founders, and infra leaders. 📍 On-site Venue (September 15–17): Santa Clara Convention Center 5001 Great America Pkwy, Santa Clara, CA 📍 On-site participation is by invitation only. Travel and accommodation expenses will not be covered.