⚡ Autonomous Software Factories + Agentic Security

⚡ Autonomous Software Factories + Agentic Security

Issue #22 · Week of April 13

Joe Heitzeberg • Founder at AI Tinkerers • ⏱️ 1 min read

Creating space for leading builders to share ideas, grow, and make an impact.

This week, we saw builders shipping concrete systems, from autonomous software factories to LLM training pipelines. Nate Aune’s Sundai Claw tackles autonomous software creation, while Sofie Van Landeghem showed how Karpathy’s Nanochat trains LLMs from scratch on a single GPU.

Several demos focused on improving developer workflows. Rohan Hasabe’s ATX introduces an automated code review loop using specialized agents. We also saw explorations into the security implications of agentic tools with Will Sergeant’s work on Agentic Trust and Security Verification, and Michael Lever’s DroneQRF platform for predictive drone security, highlighting the integration of diverse data streams.

Sundai Claw: Autonomous Software Factories Nate Aune CEO at Jazkarta 📍 AI Tinkerers - Boston • Mar 30 Cooking Read more → TECH STACK Loading tech tags... PROJECT LINKS sundai.club github.com Video @natea

Agentic Trust and Security Verification Will Sergeant Cybersecurity Analyst at Harvard Medical School 📍 AI Tinkerers - Boston • Mar 30 Cooking Read more → TECH STACK Loading tech tags... PROJECT LINKS agentichighway.ai github.com @willsarg617

ATX: Automated Code Review Loop Rohan Hasabe Software Engineer at Esper.io 📍 AI Tinkerers - Boston • Mar 30 Cooking Read more → TECH STACK Loading tech tags... PROJECT LINKS iteration.sh iteration.sh @hasaber8

DroneQRF: Predictive Drone Security Platform Michael Lever CTO at 24/7 Drone Force 📍 AI Tinkerers - Johannesburg • Mar 31 Cooking Read more → TECH STACK Loading tech tags... PROJECT LINKS

Nanochat: Train LLMs from Scratch Sofie Van Landeghem Open-source maintainer at OxyKodit 📍 AI Tinkerers - Brussels • Apr 01 Cooking Read more → TECH STACK Loading tech tags... PROJECT LINKS github.com @oxykodit

More Great Builds

Quick hits from the community — demos worth bookmarking:

TriCoach

Wafaa Arbash • AI Tinkerers - Boston • Mar 30

Wafaa Arbash presented TriCoach, a triathlon training app that serves a daily dashboard plus an AI coach that adapts to progress, needs, and effort. She built the backend with Claude Code and shipped it on Vercel, turning a Lovable prototype into something you can actually use in the browser. The demo stood out because it pairs real-world habit tracking with adaptive programming, and people seemed to really enjoy that practical feedback loop. If productized, it could scale as a personalized coaching workflow for busy athletes.

Loading tech tags...

adaptive-race-coach.vercel.app

CoChef: AI Kitchen Copilot

Jayesh Gupta • AI Tinkerers - Boston • Mar 30

Jayesh Gupta from InterSystems presented CoChef, an AI-enabled copilot in the kitchen that watches cooking in real time using a Raspberry Pi plus thermal and RGB cameras. It uses multimodal perception to produce actionable, spoken and on-screen guidance, and it lets you ask general questions while the device runs lights and speakers for feedback. Builders will appreciate the hands-on full-stack approach, from sensor inputs to conversational orchestration. We liked it because it made the vision-language gap feel practical, and people seemed especially excited about turning messy real-world scenes into reliable next steps. CoChef hints at a future where assistants ship as helpful appliances, not just chatbots.

Loading tech tags...

cochef.techVideo

Building Effective Agents

Niels Rogge • AI Tinkerers - Brussels • Apr 01

Quinten Rosseel (ML6, dotdash.ai) shared how he builds effective agents with the Claude Agents SDK, including practical lessons from turning chatty prompts into task-driven workflows. The demo runs agentic logic that can plan, act, and iterate, while the repo ai-deadlines shows the kind of automation pipeline you can inspect and fork, from YAML-driven updates to a React UI. It also landed well with folks at meetups, since it matched real developer pain around reliable orchestration and kept scope tightly shippable.

Loading tech tags...

github.com

You are one of 95,000+ readers from Apple, Amazon, Microsoft, Google, Nvidia, OpenAI, Anthropic, Cohere, ElevenLabs, Scale AI, Groq, Mistral AI, and others — spanning frontier labs, big tech, startups, and top universities.