The Patched Node Still Reports Brotli 1.2.0, and My PC Has Nine Places to Check
Node, .NET, curl and others each carry their own Brotli decoder. I counted the places on one Windows PC and tested which clients decode Brotli they never asked for.
Benchmark
Measured comparisons with reproducible numbers.
29 items
Posts
Where Do Container Image Bytes Come From? Measurements on Python Images
I unpacked Python images layer by layer and rebuilt them with common Dockerfile habits. A later rm saved nothing on the wire, and half of pandas in the image was its own test suite.
Logical Clocks Fixed the Clock-Skew Bug, but 1 Write in 200 Still Vanished
I ran wall-clock, Lamport, hybrid and vector clocks through one simulated workload. Logical clocks removed the skew mistakes, yet 0.5% of writes still vanished. Only vector clocks kept them all.
Traces Break at the Hand-Off: Five of Eleven Common Boundaries Dropped Every Request
I sent 200 requests across eleven kinds of hand-off with the OpenTelemetry Python SDK. Five lost every trace, batching lost 90%, and the failures you most want to trace were the ones left behind.
The Standard Burn-Rate Alert Pages Once Per Failed Request on a Low-Traffic Service
I simulated Google SRE Workbook burn-rate paging rules on a 99.9% SLO. They bound the budget lost before a page, but below about 900 requests a day they page for nearly every error.
A Free Speaker Filter Matched an LLM-Built Fact Layer for Agent Memory
On 1,536 LoCoMo questions, tagging each fact with its subject lifted search from 54.4% to 59.5%. Filtering the raw turns by speaker did the same, for free.
Hermes's Memory Search Beat Embeddings, but Only Through Its Last-Resort Fallback
On 1,536 LoCoMo questions, Hermes-style FTS5 session search found the answer turn in its top 10 for 54%, ahead of a small embedding model. Nearly every hit came from an OR retry.
robots.txt Allowed 87 AI Requests. The Servers Blocked 25 of Them Anyway.
I sent three AI user-agent strings to 66 sites and compared robots.txt with what the server did. Sephora allows them on paper and returns 403. BBC denies them and serves the page.
Pele Draws Mermaid Diagrams About 100x Faster, but Its Default Bundle Is Not Smaller
Pele claims to be 500x faster and 6x smaller than Mermaid.js. I ran both on 39 diagrams from this blog: the speed holds at about 100x, the size only with an opt-in import.
Four Macs Run Muse Spark at 36.7 tok/s, and a Fixed 14 ms per Token Sets the Ceiling
Apple says decoding is memory-bandwidth bound. I checked that on my own machine, then fit Apple's tables: a fixed 9 to 16 ms per token explains why more Macs help less each time.
Go 1.27 Turned 100 New Connections Into 1. Above 256 KB, Nothing Changed.
Go 1.27 now drains an unread HTTP/1 response body so the connection can be reused. I counted dials on Go 1.26 and 1.27, and found where the fix stops working.
Anthropic Says Naive Git Costs 600 PB a Day. Stock Git Cut a 1.1 GB Clone to 3.9 MB
Anthropic replaced git for its monorepo. I measured what shallow, partial and sparse checkouts do to a 343,502-commit repository, and where stock git still runs out.
A 1.4 KB Electron App Costs 125,000 Tokens to Analyze, and the Renderer Call Is Missing
I graded the REA reverse-engineering tool against four tiny Electron apps with known answers. It got the main process right, missed script written inside HTML, and returned 125k tokens.
A Hostile Claude Code Mod and a Harmless One Print the Same validate Output
claude plugin validate rejects every way I tried to hide a mods API call, and the runtime blocks raw JavaScript escapes. But its output is a capability list, not a verdict.
A One-Line Git Commit Is 334 Bytes of Data and 17,036 Bytes of Disk
A popular note says tiny Git commits cost kilobytes and big ones about 10% of the file. I repeated the tests: the kilobytes are filesystem blocks, and the 10% is only the first copy.
WSL3 Is Reported 5 to 60% Faster. On My Machine, Touching Fresh Memory Got 3.5x Slower.
A blog post says WSL 3 beats WSL 2 by 5 to 60%. I upgraded my own machine from WSL 2.4.12 and repeated its tests: syscalls and first-touch memory got slower, not faster.
Postgres Wasn't Ignoring My Index. On My Machine It Used the Index Too Often.
Why does Postgres skip indexes? On a 5M-row table the default planner picked a bitmap scan up to 1.85x slower than a sequential scan, and physical column order mattered more than selectivity.
I Pushed the For Down and My Loop Got Slower. Removing the Branch Made It 6x Faster.
I measured the performance claims behind 'push ifs up and fors down' in C. Hoisting an if is free at -O3, batching a lock saves 44x, and collecting into a batch made one loop slower.
In 2010, Skipping 15 of 16 Array Elements Saved Nothing. On My 2026 CPU It Saves Over 60%.
I re-ran Igor Ostrovsky's 2010 processor cache experiments in C on a 2026 laptop CPU. The cache cliffs survived. The famous 'skipping is free' result did not.
Red Hat's Classifier Beat Jev by 6x. My Own Attempt Just Tied It.
Red Hat benchmarked Jev against 8 other guardrail models and found a local classifier beats it on speed by 6x. I tried to reproduce that result myself, on my own task.
My Blog's Correction Ranks #15. The Mistake It Fixed Ranks #6.
liao.gg argued vector-based agent memory cannot tell correct from stale, without a benchmark. I tested the claim on this blog's own corrected numbers, with real embeddings.
It Swapped Opus 5 for Opus 5.5. The Detector Saw Nothing.
An open-source benchmark set out to catch Claude quietly getting worse after launch. Its own validation test, swapping in a different model, came back statistically indistinguishable from noise.
Jev and Gemini Agreed on Every Answer. Gemini Cost 34 Times As Much.
The same classification task, run through Jev and Gemini 3.8 Flash with real API calls: identical accuracy, four to five times the latency, eight to thirty-four times the cost.
Ten Agents Race the Same Git Ref. Nine Updates Vanish.
Ten processes update the same git ref at once. Git returns success for all ten, and by default it hides which nine of them never actually stuck.
Go's New SIMD Package Loses to a For Loop Below 256 Elements
Go 1.27 ships a portable SIMD API with zero published benchmarks. Measured on real hardware: a plain loop wins below 256 elements, then SIMD holds a steady ~2x, not the 8x its lane count implies.
Guides
RDF vs Property Graphs, Step by Step: One Fact, Two Models, and What Each Costs
A beginner-friendly comparison of RDF triples (SPARQL, OWL) and property graphs (Cypher), checked on one fact, one class hierarchy and a 1M-edge graph, with runnable code.
robots.txt Explained: How Crawlers Read It, Where Python Gets It Wrong, and Its Limits
A reference for robots.txt: the syntax, how a crawler picks a group and a rule, what happens when the file fails to load, how to write rules for AI bots, and what it cannot enforce.
How an LLM Answers One Prompt: Tokens, the KV Cache and the Decode Loop, Measured
A beginner-friendly walk through what happens between pressing send and seeing the first word, checked with a 135M-parameter NumPy model and real tokenizer counts.
Git Partial Clone and Sparse Checkout, Step by Step: What Each Flag Downloads, What Breaks
How to clone a huge repository without downloading all of it, step by step, with the measured cost of every command and the history commands that quietly download more than a full clone.