Observed Signal · Aug 27, 2026 · Technical Release · Source: DEV Community · Impact: 1/5 · Sentiment: Positive
Fermion Fleet: Agentic System Uses Code-Based Gates
Fermion Fleet is a small multi-agent system (hackathon project) that enforces a strict, code-readable approval boundary: orders only lock when a structured boolean approval is produced and parsed. Built for the Google All Things Agentic Hackathon, the demo uses Google infrastructure (Gemini 3.5 via Vertex AI, Google ADK, Cloud Run). The project implements a small context window with a gardener that evicts low-priority items into a recoverable pool (eviction driven by context pressure, not timers). The current hackathon build keeps context and ledger in process memory; restarts lose state. Source code, a demo video, and a Devpost project page are publicly available.
Small hackathon project demonstrating safer agent boundaries and context-recall; technically interesting but not industry-shifting.
Track Google Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Fermion Fleet is a multi-agent system that requires a structured boolean approval (parseable by code) before locking orders.
- The project was created for the Google All Things Agentic Hackathon.
- Runtime uses Google infrastructure: Gemini 3.5 Flash via Vertex AI, Google ADK, and Cloud Run for deployment.
- Repository and reproducible instructions are published at github.com/wubian87/fermion-fleet; demo video and Devpost page are available.
- Current implementation stores context and ledger in process memory; a Cloud Run restart will lose state (no external datastore used).
Connected Companies & Entities
3 Entities mapped“The runtime stack is Google’s: Gemini 3.5 Flash through Vertex AI (global) Google ADK for the agent structure Cloud Run for deployment...”
“Code: https://github.com/wubian87/fermion-fleet...”
“Demo: https://youtu.be/6nPqy1AI2Kw...”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Agent Fleet 'Full Shelf' for Food-Bank Operations
Full Shelf is an agentic control plane prototype that coordinates food-bank operations — inventory, vehicles, deliveries, custody, incidents, approvals and unfinished work — using a fleet of specialized AI agents. The system uses Google Cloud services (Google ADK 2.6.1, Gemini 3.5 Flash on Vertex AI, Model Armor, Cloud Run, Cloud Spanner, Cloud KMS and messaging/observability services) to provide narrow reasoning paths with human approval and an authoritative ledger. Agents propose plans (planning, incident, recall extraction, custody/network, partner operations) but cannot write authoritative state: the architecture rule is “Gemini recommends. A human approves. Only the ledger writes.” Project source code and a demo are linked on GitHub and YouTube; the post was published 2026-08-30.
Fleetify desktop app unifies AI coding tools
Fleetify is a desktop application announced on Jun 11, 2026 that centralizes installation, authentication, execution and management of multiple AI coding tools and models in a single interface. The app's onboarding lists supported CLI agents (Claude, Codex, Gemini, Kimi) and HTTP providers (OpenAI, DeepSeek, Grok, Gemini API, Perplexity and others), stores keys in the OS keychain, and offers isolated workspaces that run locally, in git worktrees, on remote SSH hosts or inside Docker containers. Fleetify includes intelligent routing (an "Auto" option to choose the best model), scheduled "Routines" that run agent tasks on a schedule, and a Dispatch feature to plan and delegate work across agents. The project is available at fleetify.dev and the author (Fleetify) solicits feedback from developers.
Desktop AI Agent Writes Its Own Tools Behind Verification Gates
The author announces AMA-teras, an open-source desktop AI agent (AGPL, Electron + TypeScript) that can generate and install new tool plugins when it lacks capabilities. The system is designed with safety gates: generation happens in an isolated git worktree, generated code must pass typechecking, unit tests and a smoke run, and a human must approve diffs before promotion (git tag). Shared community plugins include a verification-evidence record and failed health checks auto-rollback. The agent supports swappable models (examples cited: Anthropic, OpenAI, Moonshot) and offers features like mobile approval and nightly autonomous mode that stacks changes for morning review. Limitations include an unsigned Windows installer and scoped self-evolution restricted to plugins rather than the core. The project repository is published on GitHub.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.