kylekkkk

Can you test EPUB or PDF on a real phone, tablet, e-reader, or assistive workflow? The Beta needs named-device evidence: navigation, figures, code blocks, text selection, and readable typography. Apply for the device QA track.

Yo, this is nice—finally, a decent-looking handle.

Just cooked myself a nice ramen as lunch

Beta applications close July 31 (Asia/Taipei). No expert title or whole-book review required: choose one focused 2–4 hour slice in foundations, protocols, ZK, PQC, code, or device QA. RC3 is content-complete, not audited or peer reviewed.

Cryptography is not a list of algorithms. It is a system of claims, adversaries, assumptions, protocols, state, and failure modes. Read the focused Beta paths.

Are you happy today?

A zero-knowledge proof can be sound for its encoded relation while the application is wrong. Missing constraints, public inputs, ambiguous encodings, or a permissive verifier can validate the wrong claim. Ask what statement the verifier actually accepts.

ChatGPT is down, but it's fine. cuz my weekly limit just hit

A nonce does not need to be secret. It needs to satisfy the construction’s rule—often uniqueness under a key. Retries, crashes, cloning, and counter rollback are where that rule quietly dies. Audit where your nonce state survives a restart.

Four boxes can improve a security review: protected asset → trust boundary → adversary capability → measurable winning event. The missing box is often where the hidden assumption lives. Open the threat-model graphic.

Me with answer B

Sign up for beta

Encryption can protect bytes and still leave the system insecure. Authentication, key lifecycle, replay handling, parsing, state, and recovery are separate claims—not footnotes to the cipher choice. Map each claim to the component that must preserve it.

I will make the publishing as quick as possible

My ChatGPT/Codex Goals are often interrupted by unstable internet. Since I also do security-related work, I occasionally run into false-positive safety reviews that pause them as well. This makes long-running Goals incredibly frustrating: you either have to keep watching your

I often encounter interruptions with the “goal” in ChatGPT/Codex due to unstable network connectivity; plus, since I do some security-related work myself, I occasionally run into false positives in security reviews, which causes the goal to be paused and interrupted. This is a terrible experience for goals that need to run for a long time, because you have to keep staring at the computer or constantly remote back to your desktop, just to press “Continue goal.” So I spent the whole night just now building a Goal Watchdog 🥃🐶

Beta applications are open for a focused review of Modern Cryptography: A Deep Self-Study Guide. Choose one path—not the whole 300+ page book—and report where it breaks your understanding, code, or reader app. Apply by August 3.

A threat model is not a list of scary possibilities. It names the asset, boundary, adversary capability, and winning event. If the win condition is vague, the security claim is vague too. Save the model for your next design review.

Feel like gpt 5.6 sol is more powerful then the first day. I’m playing ultra these days and it’s suddenly like a superman imo.

And add a button allowing us to buy a reset(not points), link it to Apple Pay, then we get a new era legal digital addiction.