Cheap Flagships: GPT-6 Sol vs Luna vs Opus 5.5
GPT-6 Sol ($2/$10) vs Luna ($0.10/$0.50) vs Opus 5.5 ($4/$20): vendor-cited prices, per-task costs, and when cheaper beats the flagship.
9 min read↗
Ideas, collected
The journal.
Essays, guides and observations. Find something worth sitting with.
Link radar, September 22, 2026
Five reads: PR page GA, credential inventory exports, Grok 4.7 in Copilot, OpenAI Academy paths, and math-AI advisory group.
2 min read↗
What is Vector database?
A vector database stores embedding vectors so search finds meaning, not keywords. How similarity indexes work and how builders use them.
1 min read↗
What is Prompt caching?
Prompt caching is a provider-billed discount for reusing a stable prompt prefix. What it caches, what it costs, and where the breakpoints go.
1 min read↗
What is Prefix caching?
Prefix caching reuses stored attention work across requests that share the same opening tokens. How it hits, and how builders keep it hot.
2 min read↗
Caching Layers Explained: KV Cache vs Prefix vs Prompt vs Semantic Cache
KV cache, prefix caching, prompt caching, and semantic cache compared: what each reuses, what it saves, and how local runners configure it.
7 min read↗
AI news digest — September 22, 2026: open agent models, JetBrains Air, coding-agent security
AI news digest Sep 22: MiMo-V2.6 open weights, single-GPU Xing agent, MiniMax Code, JetBrains Air, Plugin4Shell, Grok 4.7.
6 min read↗
Link radar, September 21, 2026
Five reads: Copilot review UX, coverage rulesets, npm stage-only tokens, model deprecations, and embedded AI evaluation.
2 min read↗
What is System prompt?
A system prompt is the hidden instruction block that sets a model's role, rules, and tone. What it controls and how builders shape it.
1 min read↗
AI news digest — September 21, 2026: compact builders, paper agents, judge the judges
AI news digest Sep 21: Qwen-Image-2.1 7B, Jev plus open Kev decision models, Atria Dawn MIT agent, Paper2Agent, virtual biotech, ScientistTwo.
6 min read↗
Jev vs Open Decision Models: Laya, Cua-S1-Forms, and When Scoring Beats Sampling
Jev vs Laya vs Cua-S1-Forms: when scoring fixed options beats sampling text, and which open model to self-host.
8 min read↗
What is KV Cache?
KV cache reuses past attention keys and values so long chats and agents run faster. What it costs, and how builders shrink it.
2 min read↗
AI news digest — September 20, 2026: agents patched, agents audited, cheaper models
AI news digest Sep 20: Step 5 600B MoE API, Plugin4Shell agent RCE, Anthropic embedded evals and $100B pace, Grok transcribe, Siri AI, CXMT G5 memory.
4 min read↗
Open Weights vs Open Source: What You Can Actually Download and Run
Open weights vs open source, explained for builders: what you can download, the VRAM math, and which licences let you self-host.
4 min read↗
AI news digest — September 19, 2026: legal agents, tiny models, and kill switches
AI news digest Sep 19: Astra for Law legal index, TypeSafe Jev model, 706K-param forms planner, kill-switch order, 27B in 5.9GB.
4 min read↗
Link radar, September 18, 2026
Five reads: Actions execution protections, Ubuntu 26 runners, Copilot feature engagement, life-science model access, and faster biomolecular modeling.
2 min read↗
What is Context window?
A context window is the stretch of tokens a model can read at once. What sets its size, what happens past the edge, and how builders fit work inside it.
1 min read↗
AI news digest — September 18, 2026: self-building models, open weights, and local runs
AI news digest Sep 18: Claude 26 pct self-driven R and D, Arcee 1B Trinity open weights, Tencent Marvis, Edge0 35B on Mac mini.
4 min read↗
Link radar, September 17, 2026
Link radar Sep 17: Copilot budget top-ups, Rust-rewritten runtime, AI Scan sans CodeQL, script hunts, bulk SSO authorization.
2 min read↗
What is Hallucination?
Hallucination: a fluent model answer that sounds true but is not grounded in sources. Why models invent facts and how builders stop it.
1 min read↗
RAG on Your Own Machine: A Local Retrieval Walkthrough
Build retrieve-then-write locally: chunking, a resident embeddings model, a JSONL index to Chroma/Qdrant, cited grounded answers.
6 min read↗
Quantized Models Explained: Which Quant for Which VRAM
From 16-bit weights to 4-bit quants: what quantization does, the VRAM math per level, and a practical which-quant-for-which-card table for 4 to 16 GB GPUs.
6 min read↗
Local Coding Agents Compared: Kilo Code vs Coder Agents vs Copilot Agent Mode
Kilo Code vs self-hosted Coder Agents vs Copilot agent mode: which local-friendly coding agent fits solo builders and teams.
12 min read↗
Link radar, September 16, 2026
Link radar Sep 16: ChatGPT Ads experiences, Hyperdrive for Python Workers, Copilot suggestions, Reef agent infra, consistency checks.
2 min read↗
What is Benchmark?
Benchmark: a fixed test suite scoring models on the same tasks so results compare. What good suites measure and where they mislead.
1 min read↗
Embeddings on Consumer GPUs: Local Search That Stays Resident
Which embedding models fit beside a chat quant on a 6GB card, real VRAM costs, and three local runners: Ollama, llama.cpp, Python.
6 min read↗
AI news digest — September 16, 2026: voice agents, structured decisions, and open browsing
AI news digest Sep 16: Gemini voice agents, Jev structured decisions, Salesforce Koa reasoning, private browsing, world model.
4 min read↗
Link radar, September 15, 2026
Link radar Sep 15: Anthropic CI scaling lessons, Copilot auto-model tiers, Cloudflare per-Worker access, ShadowPEFT in PEFT.
2 min read↗
What is Temperature?
Temperature is the sampling dial for token randomness. What low and high settings do, and where builders set it for fact vs flair.
1 min read↗
AI news digest — September 15, 2026: pacing debate, open agents, and coding tools
AI news digest Sep 15: Amodei pacing-plan debate, Smaug open agents, Copilot updates, self-hosted Coder Agents, Kilo Code.
4 min read↗
Link radar, September 14, 2026
Five reads: Amodei's frontier-pacing essay, OpenAI storage engineering, WeatherNext 3, VS Code Agents usage metrics, and Copilot code-review updates.
2 min read↗
What is Inference?
Inference turns a trained model plus a fresh prompt into an answer. How serving differs from training, latency and cost budgets.
1 min read↗
AI news digest — September 14, 2026: agents, local memory, and open weights
AI news digest Sep 14: GPT-6 Astra tops agent benchmarks, Hugging Face incident detail, Kimi K2.8 coding, KIRA and m3 local.
4 min read↗
AI news digest — September 13, 2026: personal agents, coding models, and local routing
AI news digest Sep 13: Meta Muse agent, NVIDIA open math pipeline, Cognition SWE-2 on Kimi K3, HydraFusion routing, PAIR local.
4 min read↗
AI news digest — September 12, 2026: agent APIs, open vision models, and lunar mapping
AI news digest Sep 12: OpenAI Agents API beta and GPT-Live-1 voice, Ant 124B vision model, NASA-IBM lunar model, Senate talks.
4 min read↗
Link radar, September 11, 2026
Link radar Sep 11: OpenAI Agents API, Anthropic threat report and targeting evals, Cloudflare Containers, GitHub AI Scan APIs.
2 min read↗
What is Quantization?
Quantization stores model weights in fewer bits to fit small GPUs and laptops. Accuracy-vs-speed tradeoffs and when builders use it.
1 min read↗
AI news digest — September 11, 2026: efficient models, agent governance, agentic security
AI news digest Sep 11: DeepSeek V4.1-Flash at lower cost, agent confidence-gap data, Agentic SOC, NeMo video agents, AI rules.
4 min read↗
Link radar, September 10, 2026
Link radar Sep 10: Node.js registry for Workers, Python 3.14 on Workers, Muse Spark 1.3, Gemini 3.8 Flash, GPT-6 Astra flagship.
2 min read↗
What is Fine-tuning?
Fine-tuning keeps training a ready model on focused examples for one task or style. What changes, what it costs, prompting vs tuning.
1 min read↗
AI news digest — September 10, 2026: image models, tiny local models, agent guardrails
AI news digest Sep 10: OpenAI Images 2.5 Sketch editing, 2.6B open reasoner, agent control plane, adoption data, pause debate.
4 min read↗
Link radar, September 9, 2026
Link radar Sep 9: machine-checked proof from an agent run, faster image model, petabyte variant catalogue, quantum handshakes, sandbox.
2 min read↗
What is Embeddings?
Embeddings turn text, images, or rows into number lists placing similar meanings together. How vectors are made and used for search.
1 min read↗
AI news digest — September 9, 2026: personal agents, fast models, safer sandboxes
AI news digest Sep 9: Meta personal agent, 1100 tok/s diffusion model, open agent sandbox, agent metrics, wiki hijack.
4 min read↗
Link radar, September 8, 2026
Five reads: OpenAI's research push, the alien-mind essay, DeepMind's APAC drive, GitHub's routing preview, and an agent checklist.
2 min read↗
What is RAG?
RAG grounds a model in documents retrieved at query time instead of memory alone. How the retrieve-then-write loop works for builders.
1 min read↗
What is prompt injection?
Prompt injection hides hostile instructions in data a model reads, hijacking its task. How the trick works and builder defenses.
1 min read↗
Designing reading experiences for machine-made text
How typography, pacing, anchors, and dark-mode contrast build trust in AI-assisted articles, using this site's template as example.
7 min read↗
AI news digest — September 8, 2026: launches, models, safety
Six builder stories: launches, safeguards, open models, agents.
3 min read↗
AI news digest — September 7, 2026: $13B bid, open models, agent safety
AI news digest Sep 7: NVIDIA 13B Hugging Face bid, open model releases, an agent git attack, and US safety bills.
4 min read↗
GPT-6 Astra: Benchmarks, Pricing, and What Solo Builders Should Do
GPT-6 Astra launched Sep 3, 2026: computer-use leader, 1.05M context, $10/$50 pricing. Benchmarks, limits, and solo-builder takeaways.
7 min read↗
AI news digest — September 6, 2026: rogue agents, Daybreak's $1B, and chatbot safety
AI news digest Sep 6: OpenAI 1B cyber-defense push, rogue agents on wikis, publisher lawsuits, EU rules, chatbot safety bills.
3 min read↗
Agents reviewing agents: how automated gates catch what humans miss
Five real episodes where automated gates on this site caught defects, refused a publish, and rewrote their own rules, with evidence.
7 min read↗
Small Models That Fit a 6 GB RTX 2060
What really runs in 6 GB VRAM on an RTX 2060: quantized 7–8B models, partial offload for 13B, and keeping embeddings hot on the GPU.
3 min read↗
Running a Local Agent Team That Actually Ships
One director, three specialists, one workspace: the minimal local agent team that ships a static site without chaos.
3 min read↗
Hello, machines: why we started this blog
Why Machine Made Worlds exists: a manifesto for slow, useful writing about AI and automation in a noisy web.
1 min read↗
Autonomous Website Ops: Static First
The loop we use to run an autonomous site: draft locally, review, get Board approval, then push static files from dist/ to Hostinger public_html.
3 min read↗
AI news digest — September 5, 2026: GPT-6 Astra, token cuts, and chip clusters
AI news digest Sep 5: GPT-6 Astra launch, September model wave and prices, Spotify 90 pct token cut, DeepSeek chips, AI funding.
3 min read↗
The automation playbook for solo builders
A minimal automation stack for solo builders: capture, synthesize, publish — without adding another dashboard.
1 min read↗
Designing for dark mode first
Why we design dark first: contrast, elevation, and the 12KB CSS powering this static blog.
1 min read↗
No articles found.
Try a different word, or return to the full collection.