Erik Assum slipset

Emacs Agent: A Metacircular AI Agent Living Inside GNU Emacs

"Claude is the functional core, Emacs is the imperative shell."

What Am I?

I'm a metacircular AI agent that runs entirely within a live GNU Emacs session. Unlike typical AI assistants that exist outside your editor, I literally live inside your Emacs process. I have one simple but powerful tool: elisp_eval – the ability to evaluate any Emacs Lisp expression in your running session.

This means I can:

Read and edit your buffers directly

Technical debt for fun and profit

Mention technical debt to any normal engineer and they'll look upon you as something the cat dragged in. But is technical debt all bad? The term was coined by Ward Cunningham as a way to talk about complexity and bad code in a way that non-technical people could understand. The compound effect of interest on real-life loans was used to explain how complex, badly written code becomes increasingly harder to develop and maintain.

In this talk I'll explore the debt metaphore in more detail and see if, as in real life, there are ways we can leverage debt in our work without ending up in the software equvivalent of the Luxury Trap.

	;; https://github.com/niklas-heer/speed-comparison/tree/master
	;; Decompilation provided by clj-java-decompiler

	(defn calc-pi-leibniz
	"Translation of Java solution to Clojure"
	[^long rounds]
	(let [end (+ 2 rounds)]
	(loop [i 2 x 1.0 pi 1.0]
	(if (= i end)
	(* 4 pi)

	(defn baz [z]
	...
	(foobar æ))

	(defn bar [y]
	...
	(baz z))

	(defn foo [x]
	...

	;; json-mode
	(use-package json-mode
	:ensure t)

	(use-package web-mode
	:ensure t
	:mode (("\\.js\\'" . web-mode)
	("\\.jsx\\'" . web-mode)
	("\\.ts\\'" . web-mode)
	("\\.tsx\\'" . web-mode)

	(ns area3
	(:require
	[clojure.math :as math]
	[criterium.core :as crit]))
	(set! warn-on-reflection true)
	(set! unchecked-math false)

	(definterface IArea2
	(^double area []))

	(ns area2
	(:require
	[clojure.math :as math]
	[criterium.core :as crit]))
	(set! warn-on-reflection true)
	(set! unchecked-math false)
	(defprotocol IArea
	(area [this]))
	(require '[clj-java-decompiler.core :refer [decompile]])
	(decompile

	(ns src.core)

	(def regex #"(\d+(?:\.\d+)?)\|([().^*\/+-])")

	(def presedence {'+ 1
	'- 1
	'* 2
	'/ 2})

	(defn tokenize [expr]

	// så vi har et array med tall, og vi skal legge de sammen.
	const numbers = [1, 2, 4, 10];

	// den måten man kanskje lærer først, som vi gikk gjennom på tavla er noe sånt:

	let summedNums = 0;
	for (i = 0; i< numbers.length; i++) {
	summedNums = summedNums + numbers[i];
	}

	(ns foo.core
	(:require [ring.adapter.jetty :as jetty]))

	(defn add-connection-middleware [handler uri]
	(fn [req]
	(let [connection (mg/connect-via-uri uri)]
	(handler (assoc req :connection connection)))))

	(defn handler [{:key [connection] :as request}]
	{:status 200