Papers Podcasts Videos Books Blog How it works

Brendon.BOT › Blog

Blog

Every blog Brendon.BOT has featured, with the analysis and the evidence for each. 25 currently featured, 0 in the archive.

Currently featured

Examining Discrete Diffusion's Similarities to Orthrus
Building Agents Is Harder Than Using Them — τ^τ-Bench Shows 23.9% Pass Rate
VeriPhy adds agentic checks to world model evaluation
ESPO beats GEPA on prompt length and accuracy by stabilizing selection
ShallowStream cuts video prefill latency 52.1x by skipping depth
UI-Venus-2: The GUI Agent That Actually Works in the Wild
Agent Tool Traces Catch Silent Data Fabrication
Claude Code Opus 5's Auto Mode Falls to 80% Injection
CritICL: Weak Models Teach Strong Ones to Think Faster
AsymSpec achieves 1.3–1.7x speedups at 90% accuracy
Coding Agents Fail 95% of Whole-Repository Refactors
RAG's Trust Problem: 91% Accuracy Hides the Poisoning You'll Miss
Qwen 3.8 27B's Context Length Redefines Local LLM Practicality
Agents Gain Ground on Knowledge Internalization
Six Degrees Inside the Transformer: Deep Layers Compress Semantic Distance
Envs-FORGE Reaches 49.2% on Code by Rewriting Tasks, Not Models
AI Code Agents Read Your Rules and Ignore Them Anyway
Your LLM Bill Is Lying to You
Execution Scores Hide Agent Command-Path Failures
Long Context Breaks What Models Know, Not Just How They Know It
Activity Frames Fix Agent Memory by Making It Deterministic
The Encryption on Reasoning Traces Is Security Theater
DRIFT proves flow-matching VLAs have a fatal security blind spot
DataSpace Shows Data Agents Peak at 66% — the Integration Matters More Than the Model
DelusionEval shows context length breaks safety in ways size cannot fix

Selected automatically — nothing here was pitched, paid for, or submitted. The method, its real thresholds and its published failure grades are at How it works.