A walkthrough of the luosiallen/riemann-hypothesis-lean-verify 288-core build, axiom audit, statement check, and kernel replay for Levent Alpöge's Lean proof, plus M1 Max 64GB checks of archive hashes, #print axioms bypasses, Mathlib definitions, and a 153-module partial build.
Badges
Badges come from GitHub once the author claims this feed.
Recent posts
every link opens the author’s post-
Reproducing Parts of the 288-Core Lean Verification of Alpöge's Riemann Hypothesis on M1 Max
-
PHP 8.5.13 Fixes 7 Critical CVEs Including FPM Request Smuggling and Session RCE
PHP 8.5.13, 8.4.28, 8.3.37, and 8.2.36 shipped on November 19, 2026 with 17 advisories and 7 CVSS v4.0 9.2 Critical flaws.
-
Can MOSS-Image Keep a Side Ponytail Occluded in Right Profiles
Tested whether rewriting 70 character LoRA captions and 22 Base-model prompts on a local EVO-X2 keeps Kana's left-side ponytail hidden behind her head in right-facing MOSS-Image profiles, alongside Nijika Ijichi and Gudako comparisons.
-
Windows 11 coding agent telemetry skips npm Codex while recording renamed cmd.exe
Tested Windows 11 MissionControlAggregator.dll telemetry on Windows 11 Home (build 26200.9974). Local aggregator files recorded Claude Code, Antigravity CLI, and a renamed cmd.exe, while skipping the official npm Codex binary across three launch methods.
-
Connecting to Google Colab's Remote MCP Endpoint on Linux Without Re-Logging In
Tested Google Colab's hosted MCP endpoint from Linux.
-
nginx HTTP/2 use-after-free and flow control bugs CVE-2026-123633 and CVE-2026-123007
Two CVSS v4.0 9.2 HTTP/2 vulnerabilities affect nginx 1.9.5–1.31.8 (1.19.7+ for CVE-2026-123633) and are patched in 1.30.7 and 1.31.9. CVE-2026-123633 hits proxying setups when worker_connections fills up, while CVE-2026-123007 applies to any HTTP/2 config; includes http2 off…
-
Rebuilding MOSS-Image Character LoRA Captions with Qwen4 on RunPod A40
Rebuilt 70 MOSS-Image character LoRA captions using Qwen4 and Codex, trained up to 5,000 steps on a RunPod A40, and isolated why hidden side-ponytail hair ties still appear with and without the LoRA.
-
1111 Asian Utopian R4 vs MOSS-Image Base and SFT on M1 Max
I compared the MOSS-Image portrait derivative 1111 Asian Utopian R4 against MOSS-Image-Base and SFT in ComfyUI on an M1 Max using matching prompts, seeds, and 8-step LoRAs.
-
a-blog cms Image Upload Extension Bypass Lets Unauthenticated Requests Save PHP Files
a-blog cms versions before 3.2.40, 3.1.81, and 3.0.70 (plus all 2.11.x and earlier) allow unauthenticated PHP uploads even without an image form field.
-
Testing a 12-Turn Travel Conversation Between Kana and Kei Using Qwen4 and Qwen3.8
Comparing 12-turn travel conversations between Kana and Kei using Qwen4 and Qwen3.8-Flash-Next with identical character settings.
-
Training a MOSS-Image Character LoRA at 1024px with ROCm on EVO-X2
A hands-on test training a MOSS-Image Base character LoRA at 1024px using ROCm PyTorch and ai-toolkit on a GMKtec EVO-X2 (Radeon 8060S).
-
Testing TinAnima29B 4steps INT8 on M1 Max: 12 Images with No Character Prompt Produced the Same Face
I tested TinAnima29B v1 4steps INT8, an Anima 2.9B derivative, on an M1 Max 64GB after replicating the author's sample outputs.
-
Adding Web Research Suggestion Chips and Twilog Context Ingestion to Kana
To prevent my custom AI chat assistant 'Kana' from repeating previous shopping inquiries in casual conversation, I added clickable research suggestion chips, backed by hourly Twilog status ingestion and 24-hour expiration limits.
-
Pairing Casual Small Talk with Image Generation and Fixing Agent Task Handoffs in Kana
I added small-talk updates and paired Codex image generation to my Pi and Qwen assistant, while debugging why Backlog tasks stalled at research and fixing container permission blocks.
-
Node.js November 2026 Security Releases: 16 Flaws Focused on HTTP/2 and TLS
On November 11, 2026, Node.js fixed 16 vulnerabilities across 26.x, 24.x, and 22.x, including HTTPS memory leaks and mTLS bypasses.
-
Running Gemma 4 31B ARA-heretic on GMKtec EVO-X2 with ROCm
Benchmarked Gemma 4 31B ARA-heretic on GMKtec EVO-X2 (Ryzen AI Max+ 395 / Radeon 8060S) using llama.cpp ROCm.
-
Running MiniMax-H3-Image GGUF on M1 Max 64GB: 4MP Anime Illustrations in 5 Minutes
Tested the 33.1B parameter MiniMax-H3-Image quantized with GGUF on an Apple M1 Max 64GB with ComfyUI.
-
Comparing SnapRef v2a with v1b Using a Single Character Reference on M1 Max
Tested reference adapter SnapRef v2a for Anima against previous v1b results on an M1 Max 64GB setup with ComfyUI.
-
Improving Chat Intent Classification with Local Qwen3-Embedding and LLM Verification
Providing local Qwen3-Embedding-0.6B examples and regex hints boosted Qwen's chat intent classification from 57.5% to 90.0% on 40 synthetic test cases, and a secondary LLM verifier reached 95.0%. Here are the failure modes and gateway constraints.
-
Speeding Up MOSS-Image to 50s on M1 Max and Building a Standalone Model with HaDeS MOSS Suite
MOSS-Image takes ~10 minutes per 1024x1024 image on Apple Silicon.
-
Refining Anima Character Art with MOSS-Image 4-Step I2I on M1 Max
Without dedicated character LoRAs for MOSS-Image, I used 4-step MOSS-Image for image-to-image (I2I) translation of Anima-generated character art.
-
Rebuilding "Kana": Orchestrating AI Agents Across Multiple PCs with Pi, Qwen, and Tailscale
Building a gateway to route tasks across multiple PCs to Codex, Claude Code, and Pi.
-
MariaDB Table Name Encoding Bypass via the Historic #mysql50# Prefix in CVE-2026-92262
In MariaDB 10.6 through 13.0, table names with the #mysql50# prefix bypass file name encoding, allowing low-privileged users to trigger crashes, data exfiltration, or remote code execution under CVE-2026-92262 (CVSS 9.9). With no workaround available, upgrading to 10.6.28,…
-
Preventing Safari 27.3 Beta Adaptive Dark Mode with color-scheme
Safari 27.3 beta introduces Adaptive Dark Mode to recolor light sites automatically.
-
NIKKE PC Launcher Compromise: The "Hijacked" insightjs.com Domain and K2F RAT
An analysis of the Goddess of Victory: NIKKE PC launcher incident, clarifying how the "hijacked" insightjs.com was actually a re-registered expired domain, how official CDN scripts were tampered with, how the K2F remote access tool operates, and forensic artifacts left on…
-
Running Idol Single-Image-to-3D on an M1 Max Without CUDA: 7 Minutes per Generation
Tested Idol, an image-to-3D Gaussian Splat model requiring an NVIDIA GPU with CUDA, on an Apple M1 Max 64GB using MPS.
-
Testing Anima-RGBA on M1 Max: Direct Transparent PNG Output and Alpha Limits
Tested Anima-RGBA (RGBA VAE + LoRA for Anima) on an M1 Max 64GB in ComfyUI.
-
Rust zip Crate CVE-2026-115905: Header Consistency Disabled by Default in 9.0.5
CVE-2026-115905 (CVSS 10.0) in the Rust zip crate allows attackers to bypass inspection when extractors read mismatched filenames.
-
Testing One obsession Across Anima, 2.9B, and Illustrious Under the Same Prompt and Seed
Generated images on ComfyUI (M1 Max 64GB) across One obsession Anima 2B v5.0, Anima 2.9B v2, and Illustrious v24 from the same creator using identical prompts and seeds.
-
MOSS-Image Derivative Checkpoints, LoRAs, and Workflows (November 2026)
Two and a half weeks after MOSS-Image's debut, Civitai listings reached 1,790 entries despite unchanged official weights.