We learn, build, and share AI in the open.

WIGTN builds web products and AI systems for clients, and publishes what it learns in the open: peer-reviewed papers, released models, and reports that say what they do not show.

Our Service

Three teams: Solutions builds AI into enterprise systems, Product ships our own, and Research publishes what we find, including what went nowhere, on WIGTN Tech.

Solutions

We build it

AX for enterprises, from consulting to the systems that run it.

  • AX Consulting
  • System Integration
SK hynix logoSK hynixGlobal One ERP (i‑ERP) Platform development.With Braincrew · 2026.10
Braincrew logoBraincrewAX project partnership.Partner · 2026.10
Mind AI logoMind AIClosed beta test partner for MEGA Code.Partner · 2026.06
Product

We ship it

Open-source tools we build and use ourselves.

  • Open Source
  • Coding Agents
WIGSS cover artWIGSSDrag a component on your dev server and the source file rewrites itself.npm · Open sourceWIGTN Plugin for Claude Code cover artWIGTN Plugin for Claude CodeProduct workflows for Claude Code, with a gate on what leaves the machine.Claude Code · Open sourceWIGTN Plugin for Codex cover artWIGTN Plugin for CodexRequirements, verification and release decisions, reusable in Codex.Codex · Open source
Research

R&D

We publish it

Peer-reviewed research on LLMs, their evaluation, and speech.

  • LLM
  • Evaluation
  • STT-TTS
ACL 2026 logoACL 2026WIGVO: real-time speech translation over phone calls.System Demonstrations · STT-TTS
EMNLP 2026 logoEMNLP 2026RCPS: choosing document parsers by retrieval.Industry Track · Evaluation
NeurIPS 2026 logoNeurIPS 2026MDI: the minimum detectable improvement of an LLM judge.TAE Workshop · LLM
Work with us

Have a problem AI hasn’t solved for you yet?

Tell us where it fails today. We measure it before we build anything, bring the fix into the systems you already run, and say where the fix stops. A report you think is wrong, or an issue on one of our tools, is just as welcome.

contact@wigtn.com