August 17 2026
The better your taste, the shorter your prompts
Personal experiments in AI agents, open models, research workflows, and human use.
00 / Who
I’m Simone and I experiment with AI. Local models, agents, research pipelines, generative systems, and all that jazz.
Mostly, I’m trying to understand how these powerful tools can become usable by more people, not just developers or obsessive nerds.
AI agents are still early. They're powerful, somewhat unpredictable, fascinating, and still way too hard to use.
I’m interested in what happens when serious non-developer users can run investigations, build workflows, test claims, automate research, and create with agents as naturally as people use Google today.
This site is where I document that search.
01 / Experiments
Local models, agents, pipelines, failed attempts, technical tests.
Jul 2026
The graph beneath the conversation
Why long-running agent work needs durable state, clear handoffs, and a reliable way to resume after interruption.
Jul 2026
The system around the agent
A visual map of the cognition, execution, isolation, authority, and adversarial review systems I built around Hermes.
Jul 2026
A second brain that shuts up
How I’m teaching Hermes to absorb my judgment, take routine decisions off my desk, and interrupt me less.
May 2026
The Amazon earthworks model learned my agent's guess
An AI agent ran a fast remote-sensing research loop around a target that had never been checked carefully enough.
May 2026
How I orchestrated multiple AI systems to produce a research paper
A behind-the-scenes look at using Claude, ChatGPT, and a DGX Spark as an AI research team.
Apr 2026
How I used AI to reproduce a scientific paper
Agent loops, context collapse, multi-model peer review.
What worked, what didn't.
What worked, what didn't.
02 / Research
AI-assisted research into claims, evidence, and weird questions.
May 2026
Is Göbekli Tepe's Pillar 43 really a 13,000-year-old star map?
An independent statistical decomposition of the claim that animal carvings encode constellations and the Younger Dryas catastrophe.
Apr 2026
Are there hidden structures below the pyramids of Giza?
An independent attempt to reproduce the SAR tomography claims about deep underground structures below Giza.
03 / Notes
Short thoughts on AI, agents, workflows, and UX.
July 29 2026
Every AI agent failure I’ve looked at eventually came back to something I wrote.
July 27 2026
Delegating to an agent teaches you how badly you understood your own work.
July 24 2026
Silence from the checker is not evidence of quality. It may simply mean the checker was not good enough to find the flaws.
July 21 2026
Local models change what you ask. Privacy removes a filter you may not notice was there.
04 / Stack
A rough map of the machines, models, tools, and workflows I use to experiment with AI.
Hardware
- 2× NVIDIA DGX Spark
- Mac Studio M3 Ultra 96 GB
- Mac Mini M4 24 GB
- NVIDIA RTX 5090
Models
- Cloud: Claude, OpenAI, Grok, Gemini, Kimi
- Local / self-hosted: Qwen, Gemma, DeepSeek via Ollama, LM Studio, and LiteLLM
Agents & orchestration
- Hermes Agent
- OpenClaw, deprecated reference
- Local agent pipelines
- Multi-agent coordination
- MCP-based communication
Creative
- ComfyUI
- Flux
- Character consistency
- Image-to-video
- AI UGC pipeline
05 / Contact
I’ve spent years working between creative direction, technology, crypto, and distributed networks. Now I’m mostly interested in AI agents, research workflows, and useful experiments.
- email[email protected]
- xx.com/theotherpomp
- githubgithub.com/simossss