Welcome to Issue N.03 of Jaseci Digest, a biweekly roundup of what's happening across the Jaseci and Jac open-source ecosystem.
This issue's lead: building an AI agent is mostly scaffolding, not the agent. Building Agentic AI with Jac shows how Jac builds that scaffolding into the language, so the only code you write is the agent itself.
Featured
|
Agentic AI is easy when it's built into the language
What if building an AI agent took 10x fewer lines of code than Python? In Jac, the whole thing is just 7 lines.
| A complete AI agent |
7 lines of Jac |
|
import from .tools { web_search, read_file, write_file }
# An agent: a function the model runs, calling tools.
def research(topic: str) -> str by llm(
tools=[web_search, read_file, write_file]
);
with entry {
print(research("the best coffee in Tokyo"));
}
|
|
✗ The usual way
You write the agent and all the code around it: each tool's JSON schema, a dispatch table, the loop that runs while the model works, and the validating and retrying of its output. The same agent in Python runs to around 70 lines, and most of it is not the agent.
|
|
✓ With Jac
That supporting code is part of the language. You describe the agent, and the runtime handles the rest. The whole thing is the 7 lines above.
|
The full article explores how Jac makes building real, capable agents this simple.
|
Tutorials
The Jaseci team just kicked off a video series that teaches Jac from the ground up. Lesson 1 is a hands-on introduction: install the toolchain, write your first program, and get a feel for how the language reads. It is the first of many, with new lessons landing regularly, so subscribe to the Jaseci channel to follow along.
Community
JacHacks heads to San Francisco for a one-day, in-person hackathon built around the future of AI programming, landing July 26 at 8 AM PT at Founders, Inc. It follows a strong run for the series, with JacHacks Spring drawing 300+ builders and 81 projects online this May.
The fastest way to talk to the team and other builders. RFC chatter, debugging help, release questions, and a steady stream of show-and-tells. Worth joining whether you ship in production or are just curious.
Articles
Over one Winter semester, 75+ students in Michigan's EECS 449 shipped 17 full-stack AI applications, most of them in Jac. These were real, working products, not chatbot demos, from an adaptive scheduling assistant to a research terminal for prediction-market traders. They got there in a single term because Jac handles the interface, the logic, the data, and the AI in one language, so the teams spent their time building instead of wiring tools together.
Releases
The headline of the 0.30 series: jaclang now ships as a single self-contained jac binary, a Zig launcher with a bundled CPython. It runs without a separate Python installation or a package manager. Download the binary, add it to your PATH, and jac run and jac start work immediately. Serving and deploying are now built in as well, with the former jac-scale package folded into the binary, and Jac-native desktop apps run on that same bundled runtime. Heavier dependencies such as FastAPI, MongoDB, and Redis are installed per project, only when your configuration requires them.
|
|
Issue N.03 · July 1, 2026
|
| © 2026 Jaseci Digest. Part of the Jaseci open-source ecosystem. |
|
|