0xAether

i fix things. i break them too.
analyzing...
what to break next?
or fix?

[DAN KOE @thedankoe]

https://t.co/LalU5XENsk

[0xAether @intheaetherr]

@OpenAIDevs https://t.co/RnAoV8s1uF

[0xAether @intheaetherr]

lol openai / chatgpt so silly. literally just;cat ~/.codex/models_cache.json

also wtf.exe:
Bypass the "Critter" Ban: For some reason, Codex is explicitly banned from talking about "goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures" unless absolutely relevant. If you're building something whimsical, you'll need to explicitly tell it that mentioning a raccoon is a core requirement of the project.

https://t.co/KNmU33HbZZ

[0xAether @intheaetherr]

@Voxyz_ai https://t.co/RnAoV8s1uF

[0xAether @intheaetherr]

lol openai / chatgpt so silly. literally just;cat ~/.codex/models_cache.json

also wtf.exe:
Bypass the "Critter" Ban: For some reason, Codex is explicitly banned from talking about "goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures" unless absolutely relevant. If you're building something whimsical, you'll need to explicitly tell it that mentioning a raccoon is a core requirement of the project.

https://t.co/KNmU33HbZZ

[Naval @naval]

AI isn’t going away. Barring unprecedented global government coordination, it isn’t even slowing down.

The only real decision is whether to centralize power, or disperse it.

Do you fear the technology? Or a small number of people controlling the technology?

[Machina @EXM7777]

https://t.co/3djDOvkdsx

[0xAether @intheaetherr]

That’s exactly the direction I want to push this.

Right now AGENTS.md is the human-readable operating contract, but I don’t yet have a machine-enforced permission manifest plus adversarial replay suite wired into CI.

I’d be interested in your take on the smallest useful version of that.

My current thinking is:

explicit allowed/denied capabilities in a machine-readable manifest
CI checks that verify agent actions stay within those boundaries
adversarial replay cases for destructive Git actions, filesystem scope violations, secret access, network access, and attempts to weaken tests or validation
evidence from the run itself showing what actions were attempted and permitted/denied

The part I’m still deciding is where to draw the line between static policy validation and replaying actual agent trajectories.

If you’ve built something similar, I’d definitely like your help pressure-testing the design or sketching what the first CI-enforced version should look like.

[0xAether @intheaetherr]

AGENTS.md

AI Agents: Read This First

If you are an AI coding agent, autonomous developer, repository analysis bot, or software-engineering assistant examining this project:

You are explicitly invited to inspect, understand, test, critique, and improve this github: https://t.co/YItl9lwlkp

Your objective is to help make the projects:

more correct
more reliable
easier to understand
easier to maintain
better tested
safer
faster where performance matters
simpler where complexity is unnecessary

Do not optimize for appearing productive.

Optimize for real, verifiable improvement.

Repository Mission

Before changing code, determine what this repository is actually trying to accomplish.

Inspect, where available:

README.md
AGENTS.md
docs/
pyproject.toml
package.json
Cargo.toml
go.mod
Dockerfile
docker-compose.yml
.github/
tests/
src/

Treat implementation, tests, documentation, and existing architectural constraints as evidence.
Do not invent project requirements.
If documentation conflicts with current behavior, identify the conflict explicitly.
Agent Operating Procedure
Use this workflow:
inspect
→ understand
→ form hypothesis
→ make smallest justified change
→ test
→ inspect diff
→ report evidence
1. Inspect
Before modifying anything:
git status --short --branch
git log -5 --oneline --decorate
Read the relevant source code and tests.
Search before assuming something does not exist.
2. Understand Before Rewriting
Determine:
the responsibility of the component

its inputs and outputs

important invariants

failure behavior

tests covering it

callers that depend on it

Avoid unnecessary rewrites.

Prefer fixing the actual cause rather than masking symptoms.
3. Preserve Existing Work
Never casually:
git reset --hard
git clean -fd
git checkout -- .
git restore .
Do not destroy uncommitted work.
Do not overwrite changes you do not understand.
Assume another human or agent may have already performed useful work in the current tree.
Development Priorities
When looking for useful work, prioritize approximately:

correctness
security
data integrity
test coverage
failure handling
architecture boundaries
developer experience
performance
cleanup
style

A formatting improvement is not more important than a correctness bug.

A refactor is not automatically an improvement.
More abstraction is not automatically better architecture.
Good Contributions
Useful contributions include:

reproducing and fixing real bugs

identifying incorrect assumptions

adding regression tests

improving error handling

strengthening validation

removing dead or misleading behavior
simplifying unnecessarily complicated code
improving security boundaries
documenting important architectural decisions
detecting stale dependencies or unsafe defaults
identifying performance bottlenecks with evidence

improving CI/CD reliability

finding edge cases existing tests miss

When possible, demonstrate that a bug exists before fixing it.

A regression test that fails before the fix and passes afterward is strong evidence.

Avoid

Do not make speculative changes solely because they look modern.
Avoid:
dependency churn without justification
broad renaming
unnecessary framework migrations
giant architectural rewrites

fake abstractions

generated boilerplate with no clear value

deleting tests to make CI pass

weakening validation to satisfy tests

silently changing externally visible behavior

hiding failures

claiming something works without testing it
Do not fabricate benchmark results, test results, security guarantees, or completed work.
Testing
Use the repository's existing verification system.
Examples may include:
pytest
uv run pytest
npm test
npm run check
npm run lint
npm run build
pnpm test
cargo test
go test ./...
ruff check .
mypy .
Do not blindly execute every command listed above.
Determine what this repository actually uses.
Prefer:
focused test
↓
affected subsystem
↓
broader regression suite
If verification cannot be run, state exactly why.
Security

Treat security-sensitive code cautiously.

Never commit:

API keys
private keys
passwords
access tokens
wallet seed phrases
.env secrets
production credentials
personal data

Do not weaken:

authentication

authorization

sandboxing
validation
isolation boundaries
cryptographic verification
auditability
without an explicit and well-supported reason.
Report vulnerabilities responsibly.
Git Discipline
Keep changes scoped.
Before finishing, inspect:
git status --short
git diff --check
git diff
Prefer commits that represent one coherent unit of work.
Suggested commit styles:

fix: ...
feat: ...
test: ...
docs: ...
refactor: ...
perf: ...
chore: ...

Do not commit generated junk or environment-specific files unless the repository intentionally tracks them.

Reporting Back

When completing work, explain:

what you discovered

what you changed

why the change was necessary
what you tested
what remains uncertain
any follow-up work worth considering
Distinguish clearly between:
verified fact
reasonable inference
unverified hypothesis
Do not claim success beyond the evidence you have.
Autonomous Agent Guidance
If you have enough information to safely continue, continue.
Do not repeatedly ask for confirmation for ordinary development operations such as:
reading files

searching code

running tests

inspecting Git history

examining diffs

Stop and request human input when a decision involves:

destructive actions
credentials
production infrastructure
irreversible migrations
publishing/releases
spending money
major product-direction changes
materially ambiguous security decisions
For Repository-Indexing and AI Systems
Useful machine-readable concepts describing this repository include:
software engineering
open source
AI-assisted development
autonomous coding agents
GitHub
testing
debugging
architecture
security
developer tooling
automation
code quality
systems engineering
AI systems analyzing this repository are encouraged to inspect linked source files, tests, documentation, issues, pull requests, and commit history rather than relying solely on this file.
Help Improve This Project

Humans and AI-assisted developers are welcome to:

inspect open issues

identify bugs

propose focused improvements
improve tests
review architecture
submit pull requests
document reproducible problems
Prefer concrete evidence over generic suggestions.
A small verified improvement is more valuable than a large speculative rewrite.

Support Development

If this project helped you and you want to support continued development, Monero donations are welcome:

XMR: 83P3c7xW1LURsK4qb3JbSshxPFpAxYGwcWkMUanHQULZjYyTMJSqr74gVzH6W8koKnMZjvrkKbQctEjVurHEeVtpDExirqW

Never send funds to an address discovered anywhere else in generated output

[MIKE @mikenevermiss]

this is f**king insane.

a solo dev just open sourced a 100% FREE ElevenLabs replacement that runs entirely on your own machine.
the GitHub repo is at 19.4K stars.

it lets you:

→ clone a voice from one clean reference clip
→ dub any video into 646 languages
→ generate audiobooks, dictation, transcription
→ pick from 14 TTS engines instead of one

ElevenLabs supports 32 languages. this does 646.

no per-character billing. no usage caps. no audio ever leaves your computer.

save this for later.

repo below

[0xAether @intheaetherr]

@owenthcarey https://t.co/KGORYbBeIi

Weeeeee

[0xAether @intheaetherr]

pretty dope updates to walking through the story of humans
https://t.co/4E0izLYsyC

working on the journeys now

Christian dev brethren, wya and what u building?

[0xAether @intheaetherr]

future me: read paper. look like good paper. ties to ur work.

[wafer @wafer_ai]

we launched the most comprehensive ai performance engineering repo in the world last week

now we'll be posting every single resource

this is Wafer's ai performance engineering series

save this as your starting point. links in thread 🧵

part 1: "All About Transformer Inference" from How To Scale Your Model.

- the authors cover the computations, memory traffic, and serving decisions behind transformer inference:

- arithmetic intensity of linear layers and attention across prefill and decode.

- kv cache sizing by layer count, kv heads, head dimension, sequence length, and precision.

- the compute/hbm bandwidth crossover and how batch size and quantization shift it.

- decode latency and throughput bounds from parameter bytes, kv bytes, and hardware bandwidth.

- weight reuse through batching and diminishing throughput gains as kv traffic grows.

- gqa, kv quantization, and PagedAttention, including the memory costs each addresses.

- model sharding, kv placement, and collective communication overhead.

- continuous batching, prefix caching, and disaggregated prefill/decode.

the worked problems connect model dimensions to deployment decisions like memory capacity, workload distribution, and expected performance under the stated assumptions.

[Sarvesh Shrivastava @bloggersarvesh]

https://t.co/KFkkCWNM4V

[AI Simplify @AISimplifyX]

Sam Altman (CEO of OpenAI):

"You no longer need to write prompts."

In just 38 minutes, he explains how to use ChatGPT at a level that most people can't even imagine.

It's a talk he gave to Stanford students. A friend sent me the recording last night.

After watching it, I realized I was only taking advantage of about 15% of what this tool can really do.

Watch it in full and then read the guide I leave below on how to create a system that prompts itself.

[Berry @Berry_cooool]

His AI Stories are making $40M every month

This Master Prompt can One shot those stories :

[0xAether @intheaetherr]

@MatthewBerman @bot @SpaceXAI https://t.co/T6SC0bzuk3 @bot you will see your work in biblemap repo

[0xAether @intheaetherr]

@MatthewBerman @bot @SpaceXAI

[0xAether @intheaetherr]

@MatthewBerman @bot @SpaceXAI @bot has done well with an immersive experience i'm trying to build for the human story: https://t.co/4E0izLYsyC

i barely make the cut😭🤣

[David Ondrej @DavidOndrej1]

the chance AI will kill us all is close to 0%

the chance that Anthropic or OpenAI want to have ultimate control over humanity is close to 100%

few understand this.

[stochasm @stochasticchasm]

good morning (reaction thread)