Skip to main content
Press & Media

Delentia Labs Media Kit

Fact sheet, story angles, and media contact for journalists covering AI agents, open source, and Thailand's technology ecosystem.

41

Algorithms in the public repository

7,820

Tests Passing (public repository, CI-verified)

not claimed

Benchmark Hallucination Target

no SLA

Uptime Design SLA Target

About the Founder

IS

Ittirit Saengow (อิทธิฤทธิ์ แซ่โง้ว)

Founder & Architect, Delentia Labs · Bangkok, Thailand

Ittirit Saengow grew up in a flat community in Khlong Toei, Bangkok, studied architecture (a four-year programme that took him six and a half years), ran four businesses before this project, and does not write code. These personal facts are his own account and cannot be checked from this site; the code, releases and tests can. He conceived the FDIA, JITNA and RCT-7 ideas and builds Delentia by directing AI coding assistants and checking their work with tests. The project began on 25 June 2025. It is an early, one-person project with no outside funding and no customers yet.

Company Fact Sheet

FoundedJune 25, 2025
What it isAn open-source agent runtime (Apache 2.0) with a safety gate, human-signed approvals and a signed audit log
Installpip install delentia-os (PyPI, 2.4.0)
Algorithms41 in the public repository (a count of modules, not a performance claim)
Test Suite7,820 passed / 0 failed (public repository CI; coverage 91.37%)
Measured resultsEach with its method and limits on the Benchmark page
With a capable modelNot measured yet
Hallucination ratenot claimed (an earlier figure was withdrawn)
Customers / hosted service / SLAnone yet
Fundingnone outside
Team1: the founder, who does not write code and builds by directing AI coding assistants
LocationBangkok, Thailand
LicenseApache 2.0

Story Angles for Journalists

Human interest

"A non-coder from a Khlong Toei flat builds an open-source agent runtime with AI assistants"

Ittirit Saengow studied architecture and cannot write code. He directs AI coding assistants and checks their work with tests, which is why the project publishes its measurements, its limits and the claims it withdrew.

Technology / AI safety

"An AI agent whose risky actions wait for a human signature, with a log anyone can check"

Every action passes a gate outside the model; risky ones wait for an Ed25519 signature from a person; every episode leaves a hash-chained, signed record. A model-free demo runs in about five seconds. It has not yet been shown with a capable model.

Transparency

"A project that publishes its own corrections"

Delentia withdrew a hallucination figure, a test count and a service count it could not reproduce, and lists each correction on its own page. It states what each measurement does not show.

Research Links