Glossary · term15 of 202
attention mechanism
The transformer component that computes a model's output as a weighted sum over every token in its input, with weights set by how strongly each position matches the current query.
also written as attention · self-attention
Related terms
attention budgetThe fixed total of attention weight a model has to spend across a window, forced by softmax to sum to 1, so every added token draws share away from the others.view termsoftmaxA normalization function that converts a set of raw scores into weights that sum to 1, which is how a model divides its attention across every token in the window.view termtransformer architectureThe neural network design behind modern language models, introduced in a 2017 paper, that computes output using attention over an entire input sequence rather than processing tokens strictly in order.view term
At a glance
- cited by
- 12 posts
- categories
- 4
- first used
- apr 2026
Claude Is a Math Problem (And the Context Window Is the Equation)The Claude context window is an equation, not a chat. Why failed attempts anchor the output, where measured degradation starts, and when to clear and re-derive.
Open Source Has a Distribution Problem, Not a Building ProblemOpen source distribution decides whether your project gets found. A one-week census of 106 launches, and the title that scored 154 in one sub and 1 in another.
Token-Compression Skills Do Not Survive MeasurementToken compression skills report savings the invoice never shows. Cache reads carry 87 percent of an agent bill, and cutting tokens can push the cost up.
Micro SaaS Ideas Are Everywhere If You Stop Thinking Like a DeveloperThe founders earning $50K MRR from micro saas ideas found them by watching people work, not by browsing listicles. Observation beats brainstorming every time.
How claude skills were meant to be made (by hand)Hand-written claude code skills outperform generated ones because they carry only the delta the model needs. Drop the bloat and survive every model upgrade.
Every Model Release Ages Your SkillsClaude Code ignoring CLAUDE.md is often an aged file, not a model bug. See which instructions a release expires, and which lines to delete on day one.
WebMCP Has 20× the Interest and Zero DeploymentsWebMCP adoption is zero across 111,076 scanned domains while search interest holds near 20 times its January level. Here is what to ship now, and what to watch.
Web Scraping in the AI Age: What Actually ChangedWeb scraping in the AI age broke twice over: LLMs killed the CSS selector, and AI crawlers exploded while referral traffic collapsed. Here is what changed.
Run Claude Code From Your Phone: The 2026 Remote SetupRun Claude Code from your phone two ways in 2026: official Remote Control with a QR scan, or SSH and tmux over Tailscale for API keys and overnight runs.
Host a Minecraft or ARK Server With Claude CodeSkip the $25/month managed hosts. Use Claude Code to self-host a Minecraft or ARK game server in 15 minutes with zero DevOps knowledge required.
Vibe Coding Is Breaking Production (Here's How to Do It Right)Vibe coding shipped 1.5 million leaked API keys and got apps banned from the App Store. The problem is not AI-generated code. It is skipping the review. A practical checklist for doing it right.
Freelance Client Satisfaction: Get Rehired Every TimeBuild freelance client satisfaction that gets you rehired. Written scope, weekly updates, and over-delivery turn one-off projects into retainer income.