lukestanley

<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Glass agent v4 - Luke Stanley</title>
<script type="module" src="https://pyscript.net/releases/2026.6.1/core.js"></script>
<style>
textarea { width: 100%; box-sizing: border-box; }
</style>

lukestanley

# /// script
# dependencies = [
#   "cryptography",
#   "trio",
# ]
# ///

from csv import DictReader
from datetime import datetime, timedelta, timezone
from ipaddress import ip_address

lukestanley

#!/usr/bin/env python3
"""nano_rlm.py - a compact VM-isolated Recursive Language Model demo.

The full prompt is loaded into an isolated persistent REPL, the root model sees
only metadata, generated Python can call `rlm_query(...)` recursively, compact
observations are fed back, and setting JSON-serializable `Final` returns the
answer. The demo summarizes `rlm.pdf`.

Requires:
  - qemu-system-x86_64

lukestanley

"""
In early Nov 2025, I tried the non-instruction tuned T5-Gemma models on Colab.
Why? Modern encoder-decoder models could address prompt injection risks which right now are very high and risky.
This logs some of my successes and HF social signals. I didn't get around to finetuning it yet.
--@lukestanley
"""
# !pip install -q -U transformers bitsandbytes accelerate
from huggingface_hub import login
# login(new_session=False)
login(token=READ_ONLY_HF_TOKEN)

lukestanley

#!/usr/bin/env -S uv run
# Uses the inline script metadata Python Enhancment Proposal (PEP 723)
# To install UV see: https://docs.astral.sh/uv/getting-started/installation/
# To run with UV: uv run dspy_minimal_mcp_demo.py

# /// script
# requires-python = ">=3.11"
# dependencies = [
#   "dspy>=3.0.2",
#   "mcp>=1.2.0",

lukestanley

# Uses the inline script metadata Python Enhancment Proposal (PEP 723)
# To install UV: https://docs.astral.sh/uv/getting-started/installation/
# To run with UV: uv run share_my_wifi_qr_code.py

# /// script
# requires-python = ">=3.12"
# dependencies = ["qrcode-terminal"]
# ///

import shutil

lukestanley

lukestanley

lukestanley

《两年内 AI 如何接管一切》🚀 作者：Joshua Clymer @joshua_clymer，2025年2月7日。

`How AI Might Take Over in 2 Years` by Joshua Clymer @joshua_clymer, 7 Feb 2025.

![][image1]
我不是天生的“末日预言家”😟。但不幸的是，作为 AI 安全研究员，我的部分工作是思考那些更令人不安的情况。
I’m not a natural “doomsayer.” But unfortunately, part of my job as an AI safety researcher is to think about the more troubling scenarios.

我就像阿波罗 13 号🚀起飞前最后一刻还在忙着检查的机械师👨🔧
I’m like a mechanic scrambling last-minute checks before Apollo 13 takes off.

lukestanley

import json
import nltk
import requests
import hashlib
from diskcache import Cache
from time import sleep

# Set up disk-backed cache that caches forever
cache = Cache('disk_cache')