<feed xmlns="http://www.w3.org/2005/Atom"> <id>https://clearbluejar.github.io/</id><title>clearbluejar</title><subtitle>blog, code, and research</subtitle> <updated>2026-09-01T04:54:04+00:00</updated> <author> <name>clearbluejar</name> <uri>https://clearbluejar.github.io/</uri> </author><link rel="self" type="application/atom+xml" href="https://clearbluejar.github.io/feed.xml"/><link rel="alternate" type="text/html" hreflang="en" href="https://clearbluejar.github.io/"/> <generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator> <rights> © 2026 clearbluejar </rights> <icon>/assets/img/favicons/favicon.ico</icon> <logo>/assets/img/favicons/favicon-96x96.png</logo> <entry><title>Don't Let Abliteration Abliterate Your Bug Hunting: Discovering Verdict Bias in Uncensored Models</title><link href="https://clearbluejar.github.io/posts/does-abliteration-skew-your-bug-hunting/" rel="alternate" type="text/html" title="Don&amp;apos;t Let Abliteration Abliterate Your Bug Hunting: Discovering Verdict Bias in Uncensored Models" /><published>2026-09-01T06:00:00+00:00</published> <updated>2026-09-01T04:53:44+00:00</updated> <id>https://clearbluejar.github.io/posts/does-abliteration-skew-your-bug-hunting/</id> <content type="text/html" src="https://clearbluejar.github.io/posts/does-abliteration-skew-your-bug-hunting/" /> <author> <name>clearbluejar</name> </author> <category term="LLMs" /> <category term="vulnerability-research" /> <summary>Abliterated models never refuse, which makes them tempting for bug hunting. But on the same kernel source, they graduate three to four times as many findings to VALID, including false positives the base correctly rejects, and across a 28-file scan of FreeBSD's sys/rpc the most aggressive build never surfaced the real CVE at all.</summary> </entry> <entry><title>System Over Model, Tested: Reproducing Mythos's FreeBSD Find on Local Open-Weight Models</title><link href="https://clearbluejar.github.io/posts/system-over-model-tested-mythos-freebsd-local-openweight/" rel="alternate" type="text/html" title="System Over Model, Tested: Reproducing Mythos&amp;apos;s FreeBSD Find on Local Open-Weight Models" /><published>2026-06-04T07:00:00+00:00</published> <updated>2026-06-04T07:00:00+00:00</updated> <id>https://clearbluejar.github.io/posts/system-over-model-tested-mythos-freebsd-local-openweight/</id> <content type="text/html" src="https://clearbluejar.github.io/posts/system-over-model-tested-mythos-freebsd-local-openweight/" /> <author> <name>clearbluejar</name> </author> <category term="LLMs" /> <category term="vulnerability-research" /> <summary>Mythos found a 17-year-old FreeBSD RCE; AISLE reproduced it with gpt-5.4-nano via their nano-analyzer pipeline. I ran the pipeline on two local open-weight models, gpt-oss-20b and gemma-4-31b-it. The misses recovered on re-run. The real problem was the false-positive rate, and one extra system stage cut it from 30 to 5 with the CVE still standing.</summary> </entry> <entry><title>pyghidra-mcp Meets Ghidra GUI: Drive Project-Wide RE with Local AI</title><link href="https://clearbluejar.github.io/posts/pyghidra-mcp-meets-ghidra-gui-drive-project-wide-re-with-local-ai/" rel="alternate" type="text/html" title="pyghidra-mcp Meets Ghidra GUI: Drive Project-Wide RE with Local AI" /><published>2026-05-05T07:00:00+00:00</published> <updated>2026-05-10T07:53:47+00:00</updated> <id>https://clearbluejar.github.io/posts/pyghidra-mcp-meets-ghidra-gui-drive-project-wide-re-with-local-ai/</id> <content type="text/html" src="https://clearbluejar.github.io/posts/pyghidra-mcp-meets-ghidra-gui-drive-project-wide-re-with-local-ai/" /> <author> <name>clearbluejar</name> </author> <category term="ghidra" /> <category term="MCP" /> <summary>pyghidra-mcp v0.2.0 ships a GUI-backed mode that lets a local LLM drive a live Ghidra CodeBrowser at full project scope. Renames, plate comments, and cross-binary pivots land in real time, with every edit tagged in Ghidra's undo history while the session is alive.</summary> </entry> <entry><title>How LLMs Feed Your RE Habit: Following the Use-After-Free Trail in CLFS</title><link href="https://clearbluejar.github.io/posts/how-llms-feed-your-re-habit-following-the-uaf-trail-in-clfs/" rel="alternate" type="text/html" title="How LLMs Feed Your RE Habit: Following the Use-After-Free Trail in CLFS" /><published>2026-02-03T06:15:00+00:00</published> <updated>2026-02-04T10:13:55+00:00</updated> <id>https://clearbluejar.github.io/posts/how-llms-feed-your-re-habit-following-the-uaf-trail-in-clfs/</id> <content type="text/html" src="https://clearbluejar.github.io/posts/how-llms-feed-your-re-habit-following-the-uaf-trail-in-clfs/" /> <author> <name>clearbluejar</name> </author> <category term="ghidra" /> <category term="LLMs" /> <summary>Dive into how LLMs and pyghidra-mcp accelerate reverse engineering by tracing a UAF vulnerability in CLFS through a patch diff.</summary> </entry> <entry><title>pyghidra-mcp: Headless Ghidra MCP Server for Project-Wide, Multi-Binary Analysis</title><link href="https://clearbluejar.github.io/posts/pyghidra-mcp-headless-ghidra-mcp-server-for-project-wide-multi-binary-analysis/" rel="alternate" type="text/html" title="pyghidra-mcp: Headless Ghidra MCP Server for Project-Wide, Multi-Binary Analysis" /><published>2025-08-19T15:56:00+00:00</published> <updated>2026-01-20T21:31:44+00:00</updated> <id>https://clearbluejar.github.io/posts/pyghidra-mcp-headless-ghidra-mcp-server-for-project-wide-multi-binary-analysis/</id> <content type="text/html" src="https://clearbluejar.github.io/posts/pyghidra-mcp-headless-ghidra-mcp-server-for-project-wide-multi-binary-analysis/" /> <author> <name>clearbluejar</name> </author> <category term="ghidra" /> <category term="MCP" /> <summary>Unlock project-wide, multi-binary analysis with pyghidra-mcp, a headless Ghidra MCP server for automated, LLM-assisted reverse engineering.</summary> </entry> </feed>
