Blog

31–40 of 50 articles

  1. AI Bug Bounty Report Writing, without Invented Impact

    Draft clearer bug bounty reports with AI without inventing impact. Use an evidence packet, a report template and a human pre-submission checklist.

  2. GLM 5.3 Flash Red API: Low-Refusal GLM 5.3 Flash

    GLM 5.3 Flash Red is Z.ai's GLM 5.3 Flash modified to reduce refusals. Model ID, price, 1M-token context, tools, JSON mode and what is not verified.

  3. GhidraMCP and Claude Code for Malware Analysis

    Set up GhidraMCP with Claude Code for malware analysis in an isolated lab: install the Ghidra MCP server, point Claude Code at an LLM API, and gate write tools.

  4. Generate Suricata Rules With an LLM

    Generate Suricata rules with an LLM, then prove them with suricata -T and a pcap replay. Catch invented keywords, missing sids and noisy alerts early.

  5. Bug Bounty AI Broker: Enforce Scope Outside the Model

    Apply the broker pattern to authorized bug bounty: versioned scope, bounded tasks, fail-closed checks and evidence logs, without autonomous scanning.

  6. OpenRouter Uncensored Models vs a Security API

    OpenRouter routes to whatever uncensored models providers list. Here's when a security-focused API with abliterated models and opt-in ZDR fits better.

  7. Generate osquery Queries With an LLM

    Generate osquery queries with an LLM, then prove them in osqueryi: catch invented tables, ambiguous columns and noisy hunts before they reach your fleet.

  8. Qwen 3.8 27B Red API: Low-Refusal Qwen 3.8 27B

    Qwen 3.8 27B Red is Alibaba's mid-size Qwen 3.8 27B, modified to reduce refusals. Model ID, price, context, tools, JSON mode and known gaps.

  9. LLM Secure Code Review Without Refusals

    LLM secure code review for your own code: scope the diff, get structured findings, prove each with Semgrep, CodeQL or a failing test, and handle refusals.

  10. Generate Semgrep Rules With an LLM

    Generate Semgrep rules with an LLM, then let Semgrep decide: --validate catches schema errors, --test catches misses and false positives. Real outputs inside.