Matthew Martin matthewdeanmartin

Recommended policy compliance practices

Hachiderm's policy borderline "bot can only broadcast". So Riker Googling would probably pass, but an LLM bot like activist probably wouldn't.

Transparency

Mark account as bot in profile.
Optionally add #bot tag to each post, or other indication that it is AI generated.
If hosted on Hachiderm, add #hachybots to all posts.

This doco is mostly intended to remind the author how this stuff works, but has now been roaded tested by someone else too.

Before you start you'll need an IDE - I use Intellij with the Python plugin. You'll need to have gone through these

I manually checked these policies September 2024. All servers could vary in how they enforce their written policies.

There might be a chance?

mas.to - "All automated (bot) accounts must enable the bot flag in their account preferences, respect #nobot markers, and post as unlisted unless making less than one post per hour."
mastodon.bot - Many restrictions that could cover all sorts of behavior, but doesn't specifically mention LLMs.

Python venv:

source venv/bin/activate  # On Unix or MacOS
.\venv\Scripts\activate  # On Windows

	import re
	import unicodedata


	def strip_accents(s: str) -> str:
	# NFD: decomposes "ï" -> "i" + "¨"
	# Then we drop all combining marks (category "Mn")
	return "".join(
	c for c in unicodedata.normalize("NFD", s)
	if unicodedata.category(c) != "Mn"

	soit clone https://stackoverflow.com/q/12345 --answer 67890 --branch my-so-project
	cd my-so-project
	ls
	answer.md code0.js code1.js
	nano code1.js
	soit push
	soit branch "refactored-solution"
	# This push creates a brand new answer on the Stack Overflow post.
	soit push -m "Feat: A more performant and modern solution"
	# Pull edits made by moderator, etc

	# Update: fix the unbiased bit extractor to use disjoint pairs of independent samples,
	# which makes the 4-bit blocks uniform (so digits 0..9 after rejection are uniform).
	# Re-run the whole pipeline and tests.

	import hashlib
	import math
	from collections import Counter
	from dataclasses import dataclass
	from typing import Iterable, List, Sequence, Tuple

	import collections
	import math
	import re


	def seeded_text_key_scramble(source_text, secret_key, start_word_index, num_digits, verbose=False):
	"""
	Generates a sequence of pseudo-random digits based on a source text and a secret key.

	Args:

	# SPDX-FileCopyrightText: 2023 Hynek Schlawack <[email protected]>
	#
	# SPDX-License-Identifier: MIT
	# pylint: disable=global-statement
	from __future__ import annotations

	import atexit
	import contextvars
	import inspect
	from collections.abc import Callable, Generator

	"""
	Depends on this in conftest.py

	```python
	import os
	import pytest

	@pytest.fixture(autouse=True)
	def set_tmp_path_env(tmp_path):
	os.environ['PYTEST_TMP_PATH'] = str(tmp_path)