Platform
Spot Suite
The umbrella for my product line: shared design system, per-customer infrastructure, Stripe and cloud-marketplace billing, and a provisioner that turns a purchase into a running tenant.
Netherlands, building in public
Apps, agents, and the infrastructure they run on.
Principal Engineer and Microsoft Certified Trainer. I ship SaaS at Spot Suite, run a fleet of AI models, and keep an always-on agent on call.
Apps I build and run.
B2B SaaS on Cloudflare Workers, one product per customer environment, marketplace billing, and provisioning that stands a tenant up end to end.
xEvolve
Secure file transfer for teams replacing legacy MOVEit setups. SSO, per-tenant storage, full audit trail.
xevolve.io ↗Cloud Horizons
Multi-cloud inventory and governance with bundled modules for access, control, and guardrails.
product page ↗ClearScreen
Endpoint web-content filtering with a local agent and centrally managed policies.
clearscreen.spot-suite.com ↗Automate Certificates
Certificate lifecycle automation, from discovery to renewal, sold direct and through cloud marketplaces.
product page ↗InsureGuard AI
AI-assisted screening and intake for insurance teams.
insureguardai.com ↗AutonomyEV
Independent news and analysis on autonomy, EVs, and AI. Human-curated, globally scoped.
autonomy-ev.com ↗Plant Maxxing
A plant-care companion with social sign-in and an AI coach. Weekend build, real users.
plant-maxxing.com ↗Open source.
Tooling from the daily workflow, published as it stabilizes.
github.com/falkoro ↗no-mistakes
A push gate. git push no-mistakes runs review, lint, and tests before anything lands on a branch.
GitHub ↗mcp-offload-router
Self-hosted MCP server that routes heavy agent work to flat-rate model subscriptions instead of burning frontier tokens.
GitHub ↗agent-office
Watch Codex, Claude, OpenCode, and Antigravity work side by side in a shared browser office.
GitHub ↗hermes-agent
My fork of NousResearch's agent framework, the core of Hermes below. I contribute back upstream.
GitHub ↗mw5-remap
Visual joystick binding editor for MechWarrior 5. Because sim rigs deserve good tooling too.
GitHub ↗Agent benchmarks
Public, reproducible races between coding agents: one prompt, real cost, real wall-clock, video receipts.
bench.yair-tech.com ↗The model fleet.
Fifteen-plus providers, because I actually run them. Grouped by where they live and what they are trusted with.
Frontier, daily drivers
Claude Fable 5 / Opus 4.8- Conductor of the fleet. Architecture, judgement, review, final say.
GPT-5.5 (Codex CLI)- Large refactors with long briefs, multi-agent runs.
Grok 4.5 / Grok Heavy- Live X search, image generation, and parallel worker subagents.
Cursor- Second-opinion implementation lane.
Flat-rate coding fleet
Kimi K2.7 Code- Long-horizon agentic implementation, 256k context.
GLM-5.2- Bulk codegen and whole-repo analysis at 1M context.
MiniMax M3- Batch text work, summaries, light codegen.
Qwen 3.6 / GPT-OSS-120B / Nemotron- Review lanes and second opinions on a fixed-price pool.
Serverless and edge
Mistral Large 2411- The brain of the Hermes agent, on a flat-rate endpoint.
Cloudflare Workers AI- Answers product chat at the edge when I am offline.
NanoGPT pool- Per-bot model pools for a fleet of Discord personas.
Local, own hardware
Qwen 2.5 / Gemma (Ollama)- Small weights on my own GPUs for offline and private quick work.
The honest split: small models run at home, the heavy lifting rents its GPUs by the subscription.
An agent that never sleeps.
Hermes is my always-on agent. It runs around the clock on two boxes, a repurposed gaming laptop at home and a small VM in a datacenter, and I talk to it over Telegram from wherever I am.
It watches CI, fixes deploys, triages GitHub issues, sends me reports, and posts to my own dashboards behind Cloudflare Access. Its brain is swappable: it started on Grok, runs on Mistral Large today, and keeps Grok wired in for X search and image generation.
The open-source core is NousResearch's hermes-agent, and I contribute back upstream.
Speaking, training, or a working session.
Send a request below. You will get a confirmation email right away. Yair replies from yair@cloudevolvers.com.
This site is a work in progress. If the form misbehaves, email yair@cloudevolvers.com directly.
Let's talk apps, AI, or cloud.
Tips, collaborations, speaking, or a sharp correction. Reach out on X or LinkedIn.