Agents that build your data stack
Agents should do the work - design the pipeline, write the SQL,
watch the tables, fix the schema drift. You decide the outcome.
We build Bonacci, the agentic data platform and data-domain AI model,
plus open-source ThinkingLanguage and ThinkingMemory.
Five products, one platform for Data & AI - from a visual agentic studio to the model and language underneath it.
-
01
Bonacci Studio live flagship
Agentic data engineering. Build pipelines visually or by chatting with an AI agent.
-
02
Bonacci Flow private beta
Pure agentic. Tell it the outcome - the agents ship the pipeline.
-
03
Bonacci MoE 9B model coming soon
9B total, 2.6B active, 8 experts, top-2 routing. Purpose-built for Data Engineering & Data Science.
-
04
ThinkingLanguage live open source
A compiled, statically typed language where tables, streams and models are native types.
-
05
ThinkingMemory live open source
Agent-agnostic memory infrastructure. Layered memory, reasoning-aware retrieval.
-
↗
Intelligence Is Deflating: 938× Cheaper in Three Years
GPT-4-class capability went from $37.50 to $0.04/M tokens. The frontier held; the efficient tier collapsed. Built on Bonacci Studio.
-
↗
Is AI Taking Jobs? I Joined the Data to Find Out
Four datasets, one crosswalk. AI hasn't cut jobs — but the hiring curve is already bending along the automation line. Built in Bonacci Studio.
-
↗
The Current State of the AI Ecosystem
25 months of GitHub, arXiv, and Hugging Face data. RAG up 340% YoY. Agents right behind. Run live in Bonacci Studio.
-
↗
Outcome-Aware Agents
From token prediction to action consequence modeling. AC-LSCM paper, 36× fewer safety violations.
-
↗
Building 9B MoE Models Under $10K
How I trained three 9B MoE models for under $10K on Google TPU Research Cloud.
-
↗
All articles →
Notebook entries on data engineering, agents, and language design.
See an agent build a pipeline in 5 minutes
Open Bonacci Studio free, or book a 30-min demo and we'll show you live.
Reach us at contact@thinkingdbx.com or book time on calendly.
we read every email ✎