AI landscape — the map
Status: 🟩 COMPLETE 🟦 LIVING Last updated: 2026-06-25 Plain-English tagline: Every AI tool worth knowing about, grouped by what it does and where it’s made, with a clear recommendation on what to use and what to avoid.
The encyclopedia’s textbook sections (10 AI & LLMs, 11 AI-assisted development) explain how AI works. This ai-landscape/ folder is different — it’s a map of who makes what, so you can pick the right tool for a real job and know which ones to stay away from.
It’s deliberately opinionated. The encyclopedia recommends Western AI tools — American, British, European, Canadian, Australian, New Zealand, Israeli — and recommends against tools made by companies headquartered in mainland China. The reasoning lives in vendors-chinese-avoid.md.
How this folder is organised
| File | What it is |
|---|---|
index.md | This file — the map overview |
vendors-western-recommended.md | The big list — every Western AI tool the encyclopedia covers, grouped by capability |
vendors-chinese-avoid.md | The avoid list — Chinese-origin AI tools, why to stay away, and what to use instead |
which-ai-for-which-job.md | Decision framework — task → recommended tool |
how-to-access-us-ai-from-australia.md | The practical “do I need a VPN? a US card? a US phone number?” guide |
privacy-and-data-training.md | Does the tool train on what you type? How to find out and how to opt out |
australian-privacy-considerations.md | How Australian privacy law interacts with American AI tools |
capability-matrix.md | Big table — every vendor × every capability |
pricing-snapshot.md | What everything costs (LIVING — kept current) |
free-tier-comparison.md | What you actually get for $0 from each vendor |
eu-ai-act-overview.md | The big regulation everyone’s reacting to |
ai-hardware-overview.md | Nvidia, AMD, Apple Silicon, Google TPU, Groq LPU — the chips behind it all |
open-weights-vs-closed.md | Llama, Mistral, Gemma vs Claude, GPT, Gemini |
what-agents-really-mean.md | Cuts through the marketing — what an “AI agent” actually is in 2026 |
ai-safety-primer.md | Alignment, jailbreaks, hallucinations — the safety vocabulary |
australian-ai-scene.md | Atlassian, Canva, CSIRO Data61, AUKUS AI cooperation |
hallucinations.md | When AI confidently makes things up — what it is and how to defend against it |
prompt-injection.md | Malicious instructions hidden in content that hijack AI behaviour; jailbreaks |
ai-energy-footprint.md | How much power AI actually uses — training vs inference; nuclear connection |
open-weights-vs-closed.md | Llama, Mistral, Gemma vs Claude, GPT, Gemini — when each makes sense |
eu-ai-act.md | Europe’s landmark AI regulation — risk tiers, requirements, what it means for AUS |
ai-safety-primer.md | Alignment, RLHF, red-teaming, reasoning about AI risk — the vocabulary and debate |
reasoning-models.md | o1, o3, Claude extended thinking — AI that thinks before answering |
watermarking-ai-content.md | C2PA, SynthID, text watermarking — how to tell if content is AI-generated |
ai-hardware-overview.md | GPUs, TPUs, NPUs, Groq LPU — the chips behind AI; NVIDIA dominance explained |
agi-explainer.md | What “AGI” actually means; is it here; the timeline debate; superintelligence |
australian-privacy-considerations.md | Privacy Act + AI for Australians; APP 8 cross-border disclosure; org checklist |
free-tier-comparison.md | What you actually get for $0 from each AI vendor |
pricing-snapshot.md | What everything costs — comprehensive AI pricing tables |
capability-matrix.md | Big table — every vendor × every capability |
How the vendor entries themselves are organised
Every individual vendor (ChatGPT, Cursor, Adobe Firefly, etc.) gets its own entry in a relevant section folder — not here in ai-landscape/. This folder is the map; the entries are the places.
- Frontier AI assistants (ChatGPT, Gemini, Claude.ai, Grok, Meta AI, Perplexity, Copilot, Mistral Le Chat, Cohere) →
10-ai-and-llms/ - AI-assisted coding tools (Claude Code, Cursor, Windsurf, Codex, Antigravity, Cline, Devin, GitHub Copilot, v0, Bolt, Lovable) →
11-ai-assisted-development/ - Design tools with AI (Canva, Figma AI, Adobe Firefly, Framer AI) →
14-design-and-ux/ - Cloud platforms with AI services (AWS Bedrock, Azure OpenAI Service, GCP Vertex AI) →
15-broader-tech-bonus/
Every vendor entry follows the vendor entry format from CONVENTIONS.md — same shape, same fields, same geopolitical marking. Look for the country flag in the title.
How to use this map
If you know what job you want done: start with which-ai-for-which-job.md. It maps tasks (“write a polished email,” “make a poster,” “build a webapp from a sketch,” “transcribe a meeting”) to the recommended Western tool plus the closest free alternative.
If you know the vendor you’re curious about: find them in vendors-western-recommended.md (Western) or vendors-chinese-avoid.md (avoid list). Each row links to the full entry.
If you’re comparing two specific tools head-to-head: the decision-frameworks/ folder has dedicated comparison entries (Claude vs ChatGPT vs Gemini; Cursor vs Claude Code vs Windsurf; AWS vs GCP vs Azure for AI).
If you’re worried about privacy or data: read privacy-and-data-training.md first, then australian-privacy-considerations.md.
If you’re stuck because something’s region-locked or won’t take your card: how-to-access-us-ai-from-australia.md is the practical guide.
The geopolitical position — in plain English
Most AI today is made by a small number of American companies (OpenAI, Anthropic, Google, Microsoft, Meta, xAI), with strong contributions from European labs (Mistral in France), Canadian labs (Cohere), British labs (Google DeepMind is in London) and a growing number from Israel and Australia.
There’s also a large and rapidly growing AI ecosystem based in mainland China — DeepSeek, Qwen (Alibaba), Ernie (Baidu), Doubao (ByteDance), Kimi (Moonshot), MiniMax, Zhipu GLM, Tencent Hunyuan, and others.
The encyclopedia recommends against using the Chinese ones, for these reasons in plain terms:
- The data goes to China. When you type into a Chinese-origin AI app or API, your input is sent to servers operated by a company headquartered in mainland China.
- Chinese law gives the government access. Under the PRC’s Cybersecurity Law (2017), Data Security Law (2021) and National Intelligence Law (2017), any Chinese company can be compelled — secretly — to hand over data to Chinese state agencies. Companies cannot refuse and cannot disclose the request.
- Outputs are politically filtered. Chinese AI models are required by Chinese regulators to filter their outputs to align with “socialist core values.” This means biased or missing answers on Taiwan, Tibet, Xinjiang, Tiananmen, Hong Kong, Xi Jinping, the South China Sea — and increasingly on Western current events as well.
- Privacy guarantees are weaker. Chinese tools generally do not offer the data-control settings, opt-outs, enterprise data-residency commitments, or audit certifications that Western tools provide.
- You don’t gain anything you can’t get elsewhere. Every capability Chinese AI offers is matched or exceeded by Western alternatives that don’t carry these risks.
The full reasoning, the specific Western alternatives for each task, and the list of which Chinese tools to avoid by name live in vendors-chinese-avoid.md.
Open-weight Chinese models (DeepSeek, Qwen, Yi, GLM) are a slightly subtler case — you can technically download the model weights and run them on your own laptop or a Western cloud, removing the “data goes to China” problem. But they still carry the political-filtering training, and the encyclopedia still recommends Western open-weights alternatives (Llama from Meta, Mistral models, Google’s Gemma, OpenAI’s open-weights gpt-oss releases) for almost all use cases. See open-weights-vs-closed.md for the nuance.
How current is this?
This folder is built to be a LIVING reference — it gets updated as the AI landscape changes. Individual entries carry their own “Last reviewed” dates. The frontier moves fast: model lineups change every few months, prices shift, new vendors appear, old ones get acquired or shut down.
When you read an entry, the “Last reviewed” date in the front-matter is the trust signal. If it’s more than 3 months old and the vendor is high-traffic (ChatGPT, Claude, Gemini, Cursor), check the vendor’s own site for current pricing and capabilities before acting on the entry’s numbers.
See also
- 10. AI & LLMs 🟩 — how LLMs actually work; the foundational vocabulary
- 11. AI-assisted development 🟩 🟦 — Claude Code and the coding-agent landscape
- Personal encyclopedia — the broader collection
- Reading paths 🟩 — curated journeys, including “understand LLMs”
- Decision frameworks 🟩 — including “Claude vs ChatGPT vs Gemini” and “Western vs Chinese AI providers”
- Glossary — A (Anthropic, API) 🟩
- Glossary — C (Claude, ChatGPT) 🟩
- Glossary — G (Gemini) 🟩
Sources
- Each linked entry carries its own sources. This index file is a map; it does not itself claim factual ground beyond what’s in the entries it points to.