# Zero to Agent: Vercel x Deepmind Hackathon London: Project Gallery

- **Event:** [Zero to Agent: Vercel x Deepmind Hackathon London](https://cerebralvalley.ai/e/zero-to-agent-london)
- **When:** Sat, Mar 21 at 9:00 AM – 10:00 PM (GMT)
- **Where:** London, UK
- **Hosts:** [Google DeepMind](https://cerebralvalley.ai/u/deepmind), [Vercel](https://cerebralvalley.ai/u/vercel)
- **Projects:** 50 (6 placed)
- **Page:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery

## Projects

### 1. Signals

Research Service for Rapid Analysis of Complex Topics:
A research service designed for quick consumption of recent developments on complex subjects, delivering concise insights in text and audio formats suitable for on-the-go consumption -> walking, commuting, or during transit, for example.
How it works:
Users submit a research topic and email address through a simple web interface. The system validates user access, initiates a long-running Gemini Deep Research job, and then distills the output into a concise executive summary ready for immediate review.

- **Team:** [Oleg Zommers](https://cerebralvalley.ai/u/b2b)
- **GitHub:** https://github.com/b2b/signals
- **Demo video:** https://drive.google.com/file/d/1H3FJqNJOqyD3j3RBVIxDfkG5PZ8JoYwI/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=37

### 2. Agentnetes

Zero to a Self-Organizing AI Agency. On Demand. Our hackathon is Zero to Agent? We are taking it further. It's Kubernetes for AI Agent managed by agents. 
Type a goal. Agentnetes assembles a self-organizing team of specialist AI agents, each running in its own Firecracker microVM sandbox, that explore your codebase, write code, run tests, fix failures, and deliver together. No hardcoded roles. No sequential bottlenecks. One goal. A full AI agency. On demand. The root agent reads your repo, invents the right team, and spawns them in parallel. A provider task gets a Scout, Engineer, Tester, and Packager. A security audit gets a completely different team. Roles are fully emergent. ▎ Built on four research foundations: the RLM Pattern (MIT CSAIL) for context-in-filesystem, the AutoResearch Loop (Karpathy) for test-driven iteration, a Two-Tool MCP surface (search + execute, ~1000 tokens), and the A2A Protocol so every agent is a discoverable service.

- **Team:** [Shashi Jagtap](https://cerebralvalley.ai/u/ShashiJagtap)
- **GitHub:** https://github.com/Shashikant86/agentnetes
- **Demo video:** https://youtu.be/wYP3r-rb0_g
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=39

### 3. MoonWalkers

LLMs such as Gemini are incredibly capable, but they are bottlenecked by the traditional chat interface. Moonclaw breaks autonomous agents out of the text box and drops them into a dynamic, spatial web workspace. Simply type a prompt, and Moonclaw provisions a secure cloud computer, deploys a team of specialized AI agents, and streams their execution live to your browser. This isn't an AI conversation—it's an AI experience.

The Problem
Current AI models act as advisors. You ask them how to write a script, book a flight, or scrape a website, and they give you a list of instructions to do it yourself. If they hit an error, the loop breaks. They are trapped behind a text input.

The Solution
Moonclaw transforms the AI from an advisor into an executor. By giving the AI a cloud-based operating system, real developer tools (terminals, browsers, file systems), and a spatial UI to display its work, Moonclaw actually performs the tasks you ask for. It writes the code, runs the tests, reads the errors, and fixes them autonomously.
How we built it
We split the architecture into three distinct layers:

The Brain (Local): A local orchestrator built with CrewAI and LangGraph. It handles the planning, breaks large user prompts into smaller sub-tasks, and coordinates the agent swarm.

The Security Proxy: A middleware gateway that acts as a bouncer. It intercepts and blocks dangerous shell commands, restricts outbound domain access, and enforces hard rate limits to stop runaway loops.

The Execution Engine (Cloud): The actual work happens inside disposable Ubuntu containers running on GCP Spot instances. These are preloaded with dev utilities and headless browsers, and are destroyed immediately after the task finishes.

Observability Dashboard: Instead of trusting the AI blindly, we built a real-time UI that streams WebRTC video and WebSocket terminal logs straight to the frontend. You can sit back and watch the agents click around and type in real-time.

- **Team:** [Aman Sah](https://cerebralvalley.ai/u/amansah), [Enaiho Uwas Paul](https://cerebralvalley.ai/u/oactodev)
- **GitHub:** https://www.github.com/Oactodev/Moonclaw
- **Demo video:** https://www.youtube.com/watch?v=u3QoaT3pIMs
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=40

### 4. kratos

Watchdog is a personal Slack agent for overloaded IC (Independent Contributor) engineers. As an IC, you get bombarded with mentions all day. "this is broken", "have you fixed this yet?", "you need to look at this". and it's on you to keep track of all of it. Things slip through the cracks, people follow up, and you spend mental energy triaging instead of building.                                                                               

Watchdog watches your Slack mentions and handles them for you. It uses Gemini 3.1 Pro to triage each mention (bug, feature request, question, urgent, or noise) with a confidence score, DMs you a private summary, creates a Linear ticket with the right priority, then dispatches a coding agent that clones your repo, explores the codebase, writes a targeted fix, runs existing tests and linters, and opens a GitHub PR, all before you even read the message. If tests fail, it stops and reports back instead of shipping broken code. If the fix lands, it moves the Linear ticket to "Ready for Review" and DMs you the PR link, test results, and ticket status. You just review and merge when you're ready.

It's a multi-agent system, that turns the full loop of Slack mention -> triage -> ticket -> code fix -> tested PR -> status update into a single autonomous pipeline.

- **Team:** [Erons A](https://cerebralvalley.ai/u/erons), [Victor Aremu](https://cerebralvalley.ai/u/var)
- **GitHub:** https://github.com/Eronmmer/watchdog
- **Demo video:** https://youtu.be/6XFKK34ySqw
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=41

### 5. ARTIST

group chat inside a github repo between humans that a repo agent observes and then plans/builds; humans have Gemini-powered personal assistants taking notes.

- **Team:** [Angadh Nanjangud](https://cerebralvalley.ai/u/angadh)
- **Demo video:** https://www.loom.com/share/168b6c6b6e784b618bc6526ee6d17bdc
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=44

### 6. renaissance

ATLAS is a dual-brain agentic intelligence company. Eight AI agents with distinct roles liaise in real time over Telegram, speak with individual ElevenLabs voices, and produce both a capital investment thesis and a full GTM strategy simultaneously from the same market intelligence.
Two head agents — ATLAS (CIO) and FORGE (CMO) — each manage three specialists. SENTIMENT is the shared intelligence layer feeding both brains. Agents post to a live Telegram group, cross-reference each other's findings, and pause for human approval before anything ships.
Built on Gemini 3.1 Pro/Flash with Google Search grounding, URL Context, Code Execution, and Image generation. Persistent memory via Supabase. Streamed live to a dark three-column Next.js dashboard. One mission. Eight agents. Two outputs. Human in the loop throughout.

- **Team:** [Prince Adza](https://cerebralvalley.ai/u/madeinphantom)
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=43

### 7. SV7

It is a mobile companion when you're watching something live, and you can have it listen into the audio, such as something like a politician making sensational claims. What it does is it has an independent Google search and verification agent which goes out, checks the claims made against actual facts and sources (trusted sources, websites, attributed news websites), and gives you a review of whether this claim is true, false, or can't be verified. In the current day and age, where it's very hard to distinguish between what's true and what's not, this becomes your companion when you are looking at any sensationalised

- **Team:** [Sanchit Verma](https://cerebralvalley.ai/u/sanchitv7)
- **GitHub:** https://github.com/sanchitv7/make-it-make-sense
- **Demo video:** https://www.loom.com/share/30d65e465a9245e399dbea1734cc2930
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=45

### 8. YOLO-Youtuber

Youtube Producer automation

- **Team:** [Talha Sheikh](https://cerebralvalley.ai/u/TSForSuperman)
- **GitHub:** https://github.com/cammaro08/youtube-producer-ai
- **Demo video:** https://youtu.be/XacA_IiLBsM
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=46

### 9. LobSmash

Recreational padel runs on chaos: late feedback, misaligned partners, and plans that die in group chat. LobSmash Coach is a WhatsApp-native coach that gives timely, structured guidance (including image feedback), helps players pair up, and steers duo pre/post-match sessions—wired through Vercel, Next.js + Kapso webhooks and Gemini, with Playtomic links to book courts. It’s aimed at the everyday problem of access and alignment: cheaper than a human coach, closer than another app nobody opens.

- **Team:** [Sami Tahir](https://cerebralvalley.ai/u/trenchsheikh)
- **GitHub:** https://github.com/trenchsheikh/lobsmash-whatsapp
- **Demo video:** https://www.loom.com/share/b93e78c4113048c08675945a9a961c88
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=47

### 10. LetsBegin

LetsBegin takes a messy project idea, compiles it into a dependency graph of tasks, and coordinates who does what — AI
   agents auto-execute their tasks in the background while humans get guided through theirs one step at a time. Unlike
  chatbots that dump everything on you or coding agents that only handle code, LetsBegin manages the full project: agent
   tasks, human tasks, and hybrid handoffs where AI drafts and humans review.

- **Team:** [Yuyu Shen](https://cerebralvalley.ai/u/yuyu)
- **GitHub:** https://github.com/buildwithwhy/LetsBegin
- **Demo video:** https://youtu.be/xT5sbeAe3MQ
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=48

### 11. Ghost

Ghost is a developer memory layer for coding agents. It records a coding session, analyzes the git diff plus key screen frames, learns a developer’s recurring preferences and corrected mistakes, and turns that into repo-specific instruction files and chat context so future agent output matches how the developer actually works.

- **Team:** [hanish abdulla](https://cerebralvalley.ai/u/hanishabdulla), [Fahad Al Saud](https://cerebralvalley.ai/u/fahad_als)
- **GitHub:** https://github.com/hanishabdulla/ghost
- **Demo video:** https://youtu.be/MIZt3ppYLJk
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=49

### 12. Branch Point

A dynamic financial modeling platform that maintains a single governed base model — fed by a traceable input database — while enabling teams to branch versions, without duplicating files or losing version integrity. Updates to core assumptions automatically flow down through every branch, giving analysts full traceability from raw input to final output. It eliminates the manual reconciliation, streamlines model construction and use case deployment.

- **Team:** [Tara Davis](https://cerebralvalley.ai/u/TaraYD), [Dan C](https://cerebralvalley.ai/u/GetMeElon)
- **GitHub:** https://github.com/GetMeElon/VercelXDeepmindHacathonLondon2026
- **Demo video:** https://drive.google.com/file/d/1_BsFtZX6T-32clCtrJnUIcAi0TUp4Eyp/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=50

### 13. flashsolve

Agentic, multimodal puzzle solving.

- **Team:** [David Szalai](https://cerebralvalley.ai/u/davidscode404)
- **GitHub:** https://github.com/davidscode404/flashsolve
- **Demo video:** https://youtube.com/shorts/ORbCKlLvnWk?si=7zinnzQSGeoEQyCw
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=1

### 14. Hand Rhythm Pro

Hand Rhythm Pro is an AI-powered, browser-based rhythm game that turns your webcam into a precise musical controller. It uses MediaPipe’s real-time hand tracking to remove the need for specialised hardware, making rhythm games accessible to anyone with a laptop camera. Gemini-generated reactive visual themes and high-fidelity ElevenLabs voice feedback sit on top of this input system to deliver a polished, arcade-style experience

- **Team:** [Dave Cheng](https://cerebralvalley.ai/u/Masterasnackin), [Emilija Tashevska](https://cerebralvalley.ai/u/emily123)
- **GitHub:** https://github.com/MasteraSnackin/Hand-Rhythm-Pro-/
- **Demo video:** https://www.loom.com/share/42447ada389b4d96a17e33fe94128874
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=2

### 15. The Reflective Half

In an era where our digital interactions are fragmented across countless platforms and devices, Reflect brings everything together into one seamless, intelligent flow. It’s a unified messaging layer that accepts multimodal inputs — text, voice, images, and beyond — and curates them through an autonomous background agent designed to simplify capture while deepening reflection.

The system’s core philosophy is balance: frictionless input, deliberate reflection. Users can capture thoughts, notes, and interactions effortlessly, while the agent periodically introduces meaningful moments of friction — prompts to revisit, reframe, and learn from past inputs. Over time, this adaptive rhythm transforms scattered digital traces into structured self-growth, cultivating habits and supporting long-term memory.
Despite half of the team had to shoot halfway here it is....
At its heart, Reflect isn’t just a productivity tool; it’s a personal growth companion. It helps users not merely organize their world, but internalize insights from it — building a seamless bridge between ambient data collection and intentional personal evolution.

- **Team:** [Yu Wang](https://cerebralvalley.ai/u/Jack-W)
- **GitHub:** https://github.com/Jack-WangY/Personal-Growth-Agent
- **Demo video:** https://youtu.be/VORBXR-U0PM
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=3

### 16. VinDev

An agent that makes managing web development jobs simpler for small business owners.

- **Team:** [Brandon Wooding](https://cerebralvalley.ai/u/brandonwooding)
- **GitHub:** https://github.com/brandonwooding/zerotoagent
- **Demo video:** https://youtube.com
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=4

### 17. Patchbay

MCP Playground — a "Postman for MCP" that connects to any MCP server, auto-discovers tools, runs AI-powered diagnostic tests on every tool using Gemini, and lets you invoke tools through natural chat with a live execution graph. Built solo in one day.

- **Team:** [Aleksandar Dimov](https://cerebralvalley.ai/u/isavita)
- **GitHub:** https://github.com/isavita/patchbay
- **Demo video:** https://youtu.be/tEKPw3Je9Js
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=5

### 18. Celesto AI

Slop Fixer is an AI reviewer that hunts down visual 'slop' and UI bugs in React PRs and suggests changes to fix them. 

When you open a pull request, the bot spins up your React application, renders your configured routes across mobile and desktop viewports, and captures visual snapshots. The bot posts detailed comment with issues and a prompt to fix the bugs along with the screenshots.

See it live here - https://github.com/aniketmaurya/customer-support-agent/pull/2

- **Team:** [Aniket Maurya](https://cerebralvalley.ai/u/aniket)
- **GitHub:** https://github.com/CelestoAI/slop-fixer
- **Demo video:** https://youtu.be/9a47iCR8Vic
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=6

### 19. Zero to Consistent

Agent that helps with getting you from Zero to a Consistent Healthy Lifestyle

- **Team:** [Alok Matta](https://cerebralvalley.ai/u/alokmatta)
- **GitHub:** https://github.com/alokmatta/Zero-to-Consistent
- **Demo video:** https://zero-to-consistent.vercel.app/
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=7

### 20. Zero To World

Zero to World: upload a scan of any room, train a robot in that space.

We built a pipeline that turns a Gaussian Splat (.ply) of any real room into a working robot training simulation. Upload your scan, and within minutes Gemini 3 Flash analyses the space, detecting every table, chair, wall and obstacle via multimodal vision. Those bounds are projected into 3D coordinates and compiled into a MuJoCo physics scene that perfectly mirrors your real room. A Unitree G1 is then deployed into that simulation, using a pre-trained locomotion policy to navigate the actual layout around the real furniture, toward the real destination.

The long-term vision: every room on Earth becomes a robot training ground. No 3D artists. No hand-crafted simulators. Just scan, upload, deploy.

- **Team:** [Vikkash Sureshkumar](https://cerebralvalley.ai/u/Vikkash), [Yusuf Afifi](https://cerebralvalley.ai/u/yafifi)
- **GitHub:** https://github.com/afifi-yusuf/ZeroToWorld
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=8

### 21. Blake

Blake is an issue-routing and delegation tool that recommends the best person to own a piece of work by analyzing code ownership in a connected GitHub repository. It indexes recent commit history and long-lived ownership signals, interprets the issue’s subject matter, and uses that to identify the engineer with the strongest relevant expertise.

It solves the problem of unclear ownership in fast-moving teams, where people know what needs to be done but not who should take it. Instead of relying on intuition or tribal knowledge, Blake uses real engineering history to route work intelligently, with Linear as one of the systems it can plug into for the workflow.

- **Placement:** 3rd Place
- **Team:** [Yan Slobodianik](https://cerebralvalley.ai/u/yanslo)
- **GitHub:** https://github.com/Krumbit/Blake
- **Demo video:** https://youtu.be/YXCEaI2MAEk
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=9

### 22. AutoQuote

Salespeople in manufacturing companies lose hours Ctrl+F-ing through project documents to find product mentions buried in specs. AutoQuote connects to their ERP, understands their entire product catalog, and unleashes an agent swarm that reads every page in parallel - spotting their products, competitor products they can substitute, and generic mentions they'd otherwise miss. One click drafts the quotation back to ERP. The salesperson's job shifts from searching to deciding.

- **Team:** [Markus Meresma](https://cerebralvalley.ai/u/markusmeresma)
- **GitHub:** https://github.com/markusmeresma/autoquote
- **Demo video:** https://www.loom.com/share/d43b488959474f4d8566f3de111a1446
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=10

### 23. Byte AlShams

Mac System Health Assistant

This project proposes the development of a Mac System Health Assistant, a
lightweight local assistant that monitors system telemetry, diagnoses common
performance problems, and provides clear explanations and safe remediation
suggestions.

- **Team:** [Lucia Cipolina](https://cerebralvalley.ai/u/Kun)
- **GitHub:** https://github.com/lcipolina/mac_system_health_assistant.git
- **Demo video:** https://youtu.be/yPJDe2Ad-PI
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=11

### 24. Andrew Jenkins

UPLO is an organisational digital twin for the era of AI teammates. As AI agents join the workforce - not as tools but as team leads managing human reports — organisations need a way to steer the entire ship.

UPLO crawls your data sources (Google Drive, Slack, email, GitHub), builds a living interactive graph of your entire company, and gives leadership a directive system that cascades through both human and AI teams.

Every agent gets live organisational context via MCP so it never has to guess or hallucinate - it knows the current priorities, who's working on what, and which directives apply before it acts. Features include: 2D and 3D org visualisation, directive cascade with authority-level conflict detection, multi-model AI chat with full document context injection, communication alignment monitoring with auto-nudges, and an MCP directive server for agent compliance checking.

- **Team:** [Andrew Jenkins](https://cerebralvalley.ai/u/roonin)
- **GitHub:** https://github.com/RooJenkins/UPLO-Hackathon
- **Demo video:** https://youtu.be/IxW-y-QdxOk
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=12

### 25. Human OS

Human OS is a 3D cognitive and somatic mapping application that visualizes the human mind-body connection. It translates user inputs (thoughts, behaviors, physiological states, and task performances) into a structured knowledge graph, mapped directly onto an interactive 3D brain and body model. The Problem it Solves: Traditional mental wellness and "brain training" apps treat the mind as an isolated entity, ignoring the complex feedback loop between physical sensations and cognitive functions. Human OS bridges this gap. By identifying how specific physical states trigger mental loops (e.g., how chest tension correlates to an amygdala response and loss of executive function), it provides users with personalized, body-to-mind training protocols to rewire their cognitive architecture for peak performance.

- **Team:** [Marcus Mattus](https://cerebralvalley.ai/u/mattersmarcus)
- **GitHub:** https://github.com/marcusmattus/human-os.git
- **Demo video:** https://youtu.be/d574P3w2QEQ
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=14

### 26. Jargon AI

Jargon is an AI employee advocacy platform that helps marketing teams turn their
  employees into LinkedIn content creators. It has two AI agents:

  Research Agent — Give it your company's LinkedIn URL and it autonomously maps your
  entire competitive landscape. It scrapes your company, finds 5-6 competitors,
  identifies the most active employees at each competitor, fetches their LinkedIn
  posts, and analyzes what content formats and topics get the most engagement.
  Everything is saved as an interconnected knowledge graph that you can explore
  visually in real-time using a WebGL-powered graph (Reagraph).

  Content Agent — Uses the research data to create employee content plans grounded in
   competitor intelligence. It assigns topics to employees based on content gaps
  competitors aren't covering, sends recording request links so employees can submit
  voice memos, transcribes audio via Gemini's multimodal input, and writes LinkedIn
  posts that match each employee's natural voice using a comprehensive humanizer
  system (25 anti-AI-writing rules). It can also generate carousel images via Nano
  Banana 2 and assemble them into PDFs for LinkedIn upload.

  The platform solves the "what should our employees post?" problem by grounding
  every content decision in real competitor data — not guesswork.

- **Placement:** 1st Place
- **Team:** [Eldar Agayev](https://cerebralvalley.ai/u/eldar)
- **GitHub:** https://github.com/eldaragayev/ana-content-agent
- **Demo video:** https://www.loom.com/share/1fbf68df7dfa40629c622a3c912f7308
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=17

### 27. Egor

Alphys is an autonomous AI product manager that solves the "abandoned side project" problem — millions of projects get started with AI coding tools but die after day 3 when the developer loses direction. 

You can just copy and paste a GitHub repo URL and Alphys will run a full analysis pipeline powered by Gemini 3.1 Pro: it reads your codebase structure and scores completeness, discovers and scrapes real competitors via FireCrawl, extracts their features and pricing with Gemini's multimodal reasoning, then generates a prioritized roadmap with phased tasks, acceptance criteria, and dependencies. 

The web dashboard features an interactive canvas visualisation (project at center, competitors in orbit, "stealable" features as animated draggable nodes), a streaming AI chat for follow-up questions, and real-time updates via Supabase as the agent works. Gemini 3 Pro Image Preview generates a visual roadmap diagram, and the entire roadmap exports as structured markdown or JSON for any coding agent (Claude Code, Codex, Cursor) to execute overnight. 

A Telegram bot via Vercel Chat SDK provides mobile access to the same agent. Built with Next.js 16, Vercel AI SDK v6 (ToolLoopAgent with 5 custom tools), Supabase (PostgreSQL + Realtime), and FireCrawl. Alphys is the brain, not the hands — it decides what to build and why, so your coding agent can handle the how (and you can re-use your existing subscriptions to stuff like Claude Code, Codex, Cursor, etc.)

- **Team:** [Egor Vert](https://cerebralvalley.ai/u/egorvert)
- **GitHub:** https://github.com/egorvert/alphys
- **Demo video:** https://youtu.be/mLAz_JWcFtc
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=13

### 28. VisualBug Lite

Cursor writes your code. Sentry monitors your crashes.  Nobody watched your bugs happe,  until now.
VisualBug is Loom for bug reports. Record the screen,  skip the essay. 30 seconds to a full diagnosis, suspect  files, code fix, and GitHub issue.

Built in one day with Gemini + Vercel

- **Team:** [Himanshu Agarwal](https://cerebralvalley.ai/u/imHimanshuAg)
- **GitHub:** https://github.com/itssHimanshuAgarwal/visualbug-lite
- **Demo video:** https://www.loom.com/share/f3ac9df149624b98b343cda9c4043e87
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=25

### 29. Post First AI

Post Before You Post is an autonomous content optimizer that takes one raw idea (text, image, or video) and rewrites it into platform-native versions for X, LinkedIn, Instagram, and TikTok, visualizing your marketing metrics before you post.
Multiple agents then evaluate the options, suggest asset-specific edits, choose the best-performing version, and generate a ready-to-deploy posting strategy based on each region.

- **Team:** [Elena Jin](https://cerebralvalley.ai/u/elena), [Divya Ramesh](https://cerebralvalley.ai/u/techgoddess), [Elena M](https://cerebralvalley.ai/u/ElenaTRITON)
- **GitHub:** https://github.com/DivyaR12/postfirst
- **Demo video:** https://youtu.be/JXNB5MuZ0yc
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=15

### 30. Truth Network

The expert network market is massive, but the workflow is still manual. In a world flooded with synthetic content and noise, the scarcest resource is trusted human insight. We are building the agent layer for intelligence gathering.

Truth Network is an autonomous investigation agent that conducts real conversations with real people over WhatsApp to gather information on any topic. You send a question and a list of contacts. The agent messages each person, conducts a natural multi-turn conversation, asks for referrals, then automatically contacts those referrals building a chain of real human intelligence. 

Meanwhile, it searches the web in parallel to find additional relevant experts with their contact details. All conversations, findings, and referral chains are persisted to a database and a summary is delivered back to you.

- **Team:** [Liberatus Fusi-Akpodono](https://cerebralvalley.ai/u/Liberatus)
- **GitHub:** https://github.com/Builder-2703/truth-network
- **Demo video:** https://drive.google.com/drive/folders/1dB6IQxW0JMrBsiYJlXeIe8eON-_chhaK?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=16

### 31. Terdessa

MenuFlow AI is a Telegram-first AI application that turns any restaurant menu into a personalized digital dining experience. Users set their allergies, spice tolerance, diet, and food preferences once through a Telegram bot, then simply send a photo of a menu. The app analyzes the menu with AI, explains each dish in plain language, estimates possible allergens, spiciness, and nutrition values, generates food images, and creates both a generic shareable menu page and a personalized version tailored to the user. Friends can open the same menu and personalize it for themselves as well. The goal is to help people understand unfamiliar menus, make safer choices, and order with more confidence without relying only on literal translation or repeated questions to staff.

- **Placement:** Finalist
- **Team:** [Vlad Shuliar](https://cerebralvalley.ai/u/vbshuliar), [Mykyta Yakivets](https://cerebralvalley.ai/u/YKVTS), [Stas Tyshkovets](https://cerebralvalley.ai/u/STyshk)
- **GitHub:** https://github.com/terdessa/menuflow-ai
- **Demo video:** https://youtu.be/V-RDMSaV_NQ?si=KVhc5UG4rRNy1zvG
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=18

### 32. YourShadow.app

Shadow is an ambient AI companion that lives on your screen as an animated character. It passively listens to your meetings, detects action items in real-time, and autonomously executes them sending Slack messages, generating infographics, building and deploying landing pages, creating audio jingles, all without being asked.

No trigger words. No interruptions. You talk naturally, and Shadow just acts.

- **Placement:** Finalist
- **Team:** [Ishan Tiwari](https://cerebralvalley.ai/u/Itsish), [Alex Choi](https://cerebralvalley.ai/u/Alexchoi), [GHL Lee](https://cerebralvalley.ai/u/Jsl)
- **GitHub:** https://github.com/alexechoi/deepmind-vercel-hack
- **Demo video:** https://youtu.be/AF5PHP89jF8
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=19

### 33. Enginuity

Enginuity is an AI-powered parametric CAD application that lets users create
  3D solid models through natural language and engineering drawings — instead
  of manually operating traditional CAD software.

  The problem: Traditional CAD tools (SolidWorks, Fusion 360, etc.) have steep
  learning curves and require manual step-by-step modeling. Translating an
  engineering drawing or a verbal idea into a 3D model is slow and
  skill-intensive.

  What Enginuity does:
  - Chat-to-CAD: Describe a part in plain English ("create a flanged bracket
  with bolt holes") and an AI agent executes the full build123d modeling
  pipeline — sketch → extrude → fillet → booleans — producing a live 3D solid
  in the browser.
  - Sketch/Drawing-to-CAD: Upload or AI-generate an engineering drawing, and
  Gemini vision extracts the 3D feature tree (extrusions, holes, fillets,
  chamfers) with real dimensions, then the agent builds the model
  automatically.
  - Live feedback: Real-time 3D rendering in Three.js as each tool call
  completes, with a full agent trace panel showing the modeling pipeline as it
  executes.

  Stack: Next.js frontend, FastAPI + build123d Python backend, Gemini for
  vision/reasoning, Vercel AI SDK v6 for agentic tool calling, Three.js for 3D
  rendering.

- **Placement:** 2nd Place
- **Team:** [Rohith Raghunathan Nair](https://cerebralvalley.ai/u/rohith)
- **GitHub:** https://github.com/iamrohithrnair/zerotoagent.git
- **Demo video:** https://youtu.be/LZSYRriz5SA
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=20

### 34. SortedFelix

We’re building SortedFelix — a web/mobile app where regular people type “I need help with…” and get a personalised, short-form vertical AI video that walks them through the task step by step. For example: “I need help changing the water filter on my Sage coffee machine.” From replacing consumables in appliances to setting up tech, the app turns any how-to into a clear, visual guide — built instantly, just for you. Think iFixit, but as AI-generated short-form video.

- **Team:** [Andres Urena](https://cerebralvalley.ai/u/andresurena), [Polina Markova](https://cerebralvalley.ai/u/polpol)
- **GitHub:** https://github.com/L1N-z/sorted-felix.git
- **Demo video:** https://drive.google.com/drive/folders/1ohIqZzCm0FjB8lcGtcDraKIY0yi98I-R?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=21

### 35. CORD - Ambient intelligence screensaver agent.

https://ambient-agent.replit.app/

Every time you step away from your desk, intelligence accumulates - Slack messages, emails, calendar reminders, paper submissions, flagged anomalies. But when you return, you face a wall of notifications across fragmented surfaces.  

CORD is an ambient intelligence screensaver agent that reads everything while you're gone and briefs you the moment you're back.

I am Creative AI co-chair at NEURIPS The AI ML Conference.
The real problem: NeurIPS 2025 received 800+ paper submissions. Chairs coordinate across multiple geographies, timezones and apps - Slack, Gmail, OpenReview, shared drives. A single chair might be responsible for 50+ papers across a global team. The cognitive overhead of staying coordinated is enormous. Critical items get missed. Hallucinated citations slip through. 

https://fortune.com/2026/01/21/neurips-ai-conferences-research-papers-hallucinations/

CORD sits in the background as a screensaver. 
The Matrix-style visualisation shows live notification streams falling across surfaces - Slack, Gmail, Calendar, OpenReview. Agent cursors move across the streams in real time, reading and flagging. Notification bubbles surface key items - including a GPTZero agent running background checks on submissions, flagging hallucinated citations before they reach the proceedings.

When you return to your desk: press the mic, ask "what did I miss?" -  CORD synthesises everything with Gemini 2.5 and speaks the brief back via ElevenLabs.

"3 papers flagged with hallucinated citations. 14 submissions in your review queue. Terri needs sign-off on the travel grant doc. NeurIPS Creative AI call at 3pm."

Built for knowledge workers in high-stakes, multi-surface, multi-geography environments where attention is the scarcest resource. 

Today: academic conference coordination. 
Tomorrow: any regulated enterprise where missing one notification has consequences.


Built for knowledge workers in regulated and high-stakes environments where attention is the scarcest resource.

- **Team:** [Priya Prakash](https://cerebralvalley.ai/u/Priyascape)
- **GitHub:** https://github.com/priyascape/cord
- **Demo video:** https://www.loom.com/share/fb89e2991e244c488fafc57399bd68bf
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=22

### 36. Jonathan Team

CodeGate Guardian is an AI Security Posture Management platform for the age of agentic development. It scans GitHub repositories and AI-tool artefacts (skills, MCP configs, hooks, rules, and agent instructions), detects high-risk behaviours, and transforms raw findings into clear governance reporting.

In one workflow, teams can see risk posture, policy compliance, asset coverage, and ownership accountability across projects, helping security and engineering leaders identify issues early and drive remediation with confidence.

- **Team:** [Jonathan Santilli](https://cerebralvalley.ai/u/pachilo)
- **GitHub:** https://github.com/jonathansantilli/codegate-guardian
- **Demo video:** https://youtu.be/0o-p2R_vHok
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=23

### 37. CFLRA

Developers & Designers who implement consent solutions like a cookie banner shouldn’t need to understand & keep on track of ever-changing legal regulations. Currently they have to, and most do not exposing your company to greater legal risk.

PMs, Engineers, Designers can easily ship dark patterns which may increase your companies legal exposure in Europe, California etc.

- **Team:** [Kaylee Williams](https://cerebralvalley.ai/u/kaylee)
- **GitHub:** https://github.com/KayleeWilliams/hackathon
- **Demo video:** https://cleanshot.com/share/dvlDPG7z
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=24

### 38. Imprint

Imprint turns real-world photos into beautifully designed digital cards. Upload a photo of anything and Imprint uses Gemini vision to extract its "Design DNA" (color palette, mood, typography, texture). Then it generates a "production-quality" HTML/CSS card and deploys it to a live URL you can share instantly.

Most people have strong visual taste but no design tools to act on it. Imprint bridges that gap. One photo in, one stunning card out.

Built solo by a product designer using Claude Code as an AI orchestrator managing parallel subagents across git worktrees.

- **Team:** [Sean Smith](https://cerebralvalley.ai/u/SeanSmith)
- **GitHub:** https://github.com/SeanSmithDesign/Imprint
- **Demo video:** https://www.loom.com/share/7b6223f856b345e28500caff180647da
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=26

### 39. Scholes AI

A mathematically accurate technical report writing agent, that creates charts, writes proofs, as well as has a built in compiler to self-validate everything in one place. This removes the need for quants to run the papers from scratch and allows for everyone even those who are not mathematically inclined to still be able to research and test markets and mulit-million $ assets using Latex to convert to PDF.

- **Team:** [Raja Nabeel](https://cerebralvalley.ai/u/Raja820), [doudou H](https://cerebralvalley.ai/u/mame), [Alex Berard](https://cerebralvalley.ai/u/cautcoed)
- **GitHub:** https://github.com/CautCode/scholes-ai
- **Demo video:** https://youtu.be/sP4K6qTzwg4
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=27

### 40. Duel LM

If ARC-AGI-3 and LMArena had a baby. LLMs compete in duels across a myriad of turn based games in an attempt to achieve the highest ELO.

- **Team:** [Samir Lasseur](https://cerebralvalley.ai/u/samirlasseur)
- **GitHub:** https://github.com/samlasseur/Vercel-DeepMind-new
- **Demo video:** https://youtu.be/p7p6xPIIU7w
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=28

### 41. mumbl

mumbl is a multimodal AI agent orchestration platform that transforms messy, spoken stream-of-consciousness into structured, real-world actions. Users speak naturally — "I was up too late, need to move my gym session, find somewhere healthy for dinner, and draft that business proposal" — and Mumbl. interprets the intent, dispatches 7 specialized AI agents in parallel (Email, Calendar, Maps, Notes, Nutrition, Context, Website), and delivers actionable outcomes: emails drafted, restaurants found with map embeds, documents generated in Notion, calendar events rescheduled, and more. Built with Next.js 16, Gemini 2.5 Flash with function calling for agent orchestration, ElevenLabs for speech-to-text and text-to-speech, Google Places API for real venue search, and Notion API for live document creation. The tagline: "Say it broken. Get it done."

- **Team:** [Grigory Dabizha](https://cerebralvalley.ai/u/GrigoryDabizha), [Archit Bhardwaj](https://cerebralvalley.ai/u/architbhardwaj)
- **GitHub:** https://github.com/GrigoryDabizha/mumbl
- **Demo video:** https://youtu.be/oMcaskItmS8
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=29

### 42. ProcrastiNope

Our application targets individuals with ADHD or those who struggle with task planning. We developed our application ProcrastriNope based on theories of attention and productivity, our app effectively enhances user focus—specifically addressing the 'initiation deficit' (struggle to get started) often faced by those with ADHD.

- **Team:** [Andreas Stathopoulos](https://cerebralvalley.ai/u/statho7), [Fandi Meng](https://cerebralvalley.ai/u/Fandi)
- **GitHub:** https://github.com/statho7/ProcrastiNope
- **Demo video:** https://www.loom.com/share/c88a02e0efbd43c583fb7e24dd78ebf8
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=30

### 43. Lighthouse

Lighthouse is an AI-powered deal qualification engine for Vercel enterprise sales. Paste any website URL and in under 60 seconds it returns a complete sales package: tech stack detection, Core Web Vitals audit, deal score (0-100), migration strategy, architecture diagrams, PoC proposal, and audience-targeted talking points. Built with Next.js 15, Gemini 2.5 Pro, and Cortex graph memory, every analysis is persisted so the system learns migration patterns across prospects.

- **Team:** [Mike Darlington](https://cerebralvalley.ai/u/DarlingtonDev), [Parul Sinha](https://cerebralvalley.ai/u/parul), [Jude Nimo](https://cerebralvalley.ai/u/judenimo)
- **GitHub:** https://github.com/DarlingtonDeveloper/lighthouse
- **Demo video:** https://www.loom.com/share/e4ffd7460e3b44cdb3b04b74602b4e0c
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=31

### 44. gonzo

Gonzo Bot is a Discord-based AI travel agent that generates personalised day plans across three "gonzo" axes — Bourdain (food & culture), Pilkington (weird & obscure), and Hangover (urban chaos) —
  using real-time web search, Google Maps verification, AI-generated scene images, and custom character voice whispers. It solves the problem that 67% of millennials worry they aren't having authentic
   travel experiences, while existing AI travel tools just regurgitate the same "top 10" lists — Gonzo Bot finds genuinely local, currently-open places, learns your taste from reactions over time, and
   sequences everything into a weather-aware, proximity-optimised day arc with built-in dead time for spontaneous encounters. perfect when planning with your group.

- **Team:** [Sanay Shah](https://cerebralvalley.ai/u/sanay)
- **GitHub:** https://github.com/Sanyo4/vercel-deepmind-hackathon
- **Demo video:** https://drive.google.com/file/d/1Vkl8OefpOrsxrZ7m9mWTsbd-TOl_bI4L/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=34

### 45. SoloDeveloper

Franchise tracking tool where Gemini 3 populates title information and collects and updates news regularly. Tool helps to track favourite series with exact release dates and links to official sources. Nano Banana creates cover images

- **Team:** [Valentina Biriukova](https://cerebralvalley.ai/u/ValVal)
- **GitHub:** https://github.com/MyLtYkRiTiK/ARCTrack
- **Demo video:** https://drive.google.com/file/d/1lF8nZ1n1vtIqfx9kRSzT6r1CztjCg9_t/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=32

### 46. Silvester Muir Labs

Traffic incident liability analyzer. Input dash cam footage of an incident and the app will analyze the liability of each driver involve, providing full reasoning, timelines and relevant timestamps. This improves the efficiency of insurance claim processes, saving a huge amount of manual human analysis that is currently required.

- **Team:** [Samuel Silvester](https://cerebralvalley.ai/u/sjtsilvester), [Imogen Muir](https://cerebralvalley.ai/u/imogenmuir)
- **GitHub:** https://github.com/sjtsilvester/vercel-deepmind-hackathon and https://github.com/imoshimuir/collision-forensics
- **Demo video:** https://drive.google.com/file/d/1tY92CrhQpe1gHtYNpBHAZEMpmGoHjq5W/view?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=33

### 47. ZC

Lore Forge is a Discord bot that gives communities a persistent shared memory. Members submit knowledge entries via /submit, Gemini validates them for consistency against everything already established and flags contradictions, and a maintainer approves with a corresponding reaction. The result is a self-building knowledge base that anyone can query in plain English with /ask — solving the universal problem of communities repeatedly losing decisions, lore, and context to Discord's chat history.

- **Team:** [Earvin Zachary Cabrera](https://cerebralvalley.ai/u/zc954)
- **GitHub:** https://github.com/ZC-954/Lore-Forge-Hackathon.git
- **Demo video:** https://youtu.be/DsVTlxkAD0c
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=35

### 48. Lens

Lens is a production-ready, multi-modal autonomous DevOps agent that lives in Slack. 

When a team member posts a bug report or an error screenshot, Lens springs into action. Using a Human-in-the-Loop (HITL) architecture, it acts as an AI site reliability engineer: 

1. Visual Analysis: It uses Gemini's vision capabilities to extract error details and stack traces directly from uploaded screenshots.

2. Production Verification: It autonomously queries Vercel's live request logs via API to correlate the reported issue with actual 404s or 500s happening in production.

3. Repository Exploration: It explores the connected GitHub repository to fetch the offending source code.

4. Validation Sandbox: It drafts a fix and runs it through an ephemeral TypeScript compiler sandbox to ensure there are no syntax or type errors before proposing it.

5. Glass Box Telemetry: The entire reasoning loop, tool calls, and execution timeline are streamed live to a futuristic "Neon Telemetry" Next.js dashboard. 

6. 1-Click Remediation: A human reviews the fix on the dashboard or in Slack, clicks "Approve", and Lens automatically opens a GitHub Pull Request with the validated patch.

- **Team:** [Jaume Noguera](https://cerebralvalley.ai/u/jnd0)
- **GitHub:** https://github.com/jnd0/lens
- **Demo video:** https://youtu.be/lpRr04SzP5A
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=38

### 49. Ube.ai

The Autonomous Reasoning Engine for Cross-Cultural Music Curation
Ube is a next-generation "Music Strategist" built to solve the "Cold Start" problem of complex human identities. Instead of simply generating a playlist based on keywords, it uses Constrained Reasoning to bridge the gap between fragmented cultural histories and specific event narratives.

- **Placement:** Finalist
- **Team:** [Charlotte Gong](https://cerebralvalley.ai/u/CGC)
- **GitHub:** https://github.com/CharlotteGong65/Ube
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=42

### 50. GTC (Go-to-China) agent

Hardware founders visiting Shenzhen face a fragmented, overwhelming landscape: critical knowledge is buried across outdated guides, gated WeChat groups, and personal networks. No tool actually acts for them. GTC Agent changes that. It's not a chatbot that gives advice, it's an agent that executes multi-step workflows on your behalf:

1. Action engine: drafts bilingual supplier outreach messages (English + Chinese), generates personalized pre-trip checklists, parses BOMs and suggests sourcing options, and tracks production timelines with reminders
2. Curated knowledge base: fine-tuned with RAG on first-hand experience from a 2-month China hardware residency, with vetted factory contacts across categories (motors, CNC, PCB, casting, batteries), sourcing platforms, and on-the-ground navigation tips. The knowledge base grows with every conversation: user tips, factory reviews, and sourcing insights are captured, verified, and folded back in, so the agent gets smarter the more founders use it.
3. Network connector: matches founders to verified suppliers, makerspaces, and key ecosystem contacts based on what they're building

The agent reasons about each founder's experience level, project type, and journey phase to adapt its responses. A first-timer gets step-by-step logistics guidance, a repeat visitor gets straight to factory intros.

- **Team:** [Miyu Horiuchi](https://cerebralvalley.ai/u/miyuh)
- **GitHub:** https://github.com/miyu-horiuchi/gtc-agent
- **Demo video:** https://drive.google.com/drive/folders/1kkHpUMzQs7-H-uC1sn75SyL1NDgVtjFf?usp=sharing
- **Project:** https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery?project=36

---

Markdown version of https://cerebralvalley.ai/e/zero-to-agent-london/hackathon/gallery. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt