Page 3 of 50
AI automation insights
Practical guides for Australian businesses navigating AI automation. No jargon, no hype — just what works.
August 2026 • 6 min read • AI Strategy
NVIDIA and Microsoft Are Lobbying for Open Weights. Read Their Interests Before You Change Strategy
NVIDIA and Microsoft want open-weight AI. Before an Australian business treats their open letter as strategy, check whose interests it serves.
August 2026 • 7 min read • AI Strategy
Kimi K3's 2.8-Trillion-Parameter Weights Just Dropped. Here Is What Australian Businesses Should Do Next
Kimi K3's 2.8T open-weight release is a real engineering result, but most Australian SMBs don't need to react. Here's the checklist we're using this week.
August 2026 • 7 min read • ROI & Business Case
When an Open Model Needs 64 Accelerators: The New Serving Floor for Frontier Open Weights
Frontier open-weight models need 64+ accelerators to serve. Here's the real Australian cost, and why Claude often wins on price.
August 2026 • 7 min read • ROI & Business Case
The Self-Hosting Break-Even Is a Token Number. Here Is the Maths for Australian SMBs
The self-hosting versus Claude subscription decision comes down to one number: monthly token volume. Here is the Australian SMB maths.
August 2026 • 7 min read • Industry Guide
AI for Australian Insurance Brokers: Where Claude Fits and Where Open Models Do Not
Insurance broking runs on documents. Here is where Claude fits a broking practice, where a self-hosted open model actually earns its place, and how to start.
August 2026 • 6 min read • Technical
Build a Hallucination Eval Harness Before You Trust Any Open-Weight Benchmark
Vendor benchmarks won't catch hallucinations in your workflow. Here's how to build your own eval harness in a week before trusting any open-weight model.
August 2026 • 7 min read • Industry Guide
AI for Australian Recruitment and Staffing Agencies: A Claude-First Playbook
How Sydney and Brisbane recruitment agencies can use Claude to cut admin hours without crossing Privacy Act or anti-discrimination lines.
August 2026 • 7 min read • Industry Guide
AI for Australian Mining Services and Resources Contractors: A Practical Guide
Where Claude actually earns its keep for drilling, geotechnical and maintenance contractors: tender writing, safety documentation and the architecture that respects Australian site constraints.
August 2026 • 7 min read • Technical
Serving a 2.8-Trillion-Parameter MoE: Expert Parallelism, KV Cache and What a 1M Context Really Costs
Frontier open-weight MoE models need dozens of accelerators, not one. Here's why expert parallelism and KV cache costs blow out self-hosting budgets for AU teams.
August 2026 • 6 min read • AI Strategy
Claude Opus 5 Is Here: Frontier Performance at Half the Cost
Claude Opus 5 landed 24 July 2026 as the new default on Claude Max, matching near-frontier results at roughly half the cost. Here's what it means for AU Claude spend.
August 2026 • 6 min read • ROI & Business Case
Claude Model Guide: Which Model Actually Fits Your AU Business
Anthropic's new model-picking guide says start with the most capable model, not the cheapest. Here is what that means for AU businesses running Claude.
August 2026 • 7 min read • Technical
Claude Cut 80% of Its System Prompt: What That Means for Your Claude Code Setup
Anthropic trimmed over 80% of Claude Code's system prompt for Claude 5-generation models with no drop in results. Here's what it means for your CLAUDE.md file.
August 2026 • 7 min read • Industry Guide
Claude vs OpenAI in Public Research: What OpenAI's $17M Genesis Mission Signals for AU Institutions
OpenAI's $17M Genesis Mission funds US public research. Here's what it means for Claude and AU universities, agencies, and research institutes weighing a deployment.
August 2026 • 6 min read • AI Strategy
Claude Just Made Partner Certification a Four-Tier System: Why AU Firms Should Care
Claude's certification program just went from one credential to four. Here's what the new tiers mean for Australian firms buying AI automation services.
August 2026 • 6 min read • Industry Guide
Claude Design in Practice: How One Anthropic Designer Prototypes Without a Design Tool
An Anthropic designer's account of building Claude Design shows Australian SMBs a cheaper way to test ideas before paying for mockups.
August 2026 • 6 min read • ROI & Business Case
Claude's Voice Mode Now Runs on Opus and Sonnet
Claude's voice mode now runs on Opus and Sonnet, connects to Gmail and Slack, and speaks 11 languages. Here's what it's worth to a Sydney business, in dollar terms.
August 2026 • 6 min read • Industry Guide
What an AI-Powered Cyber Investigator Built on Claude Code Means for Australian Businesses
Outtake built an autonomous cyber investigator on Claude Code. Here's what the case study actually shows, and what it means for how Australian SMBs think about security monitoring.
August 2026 • 6 min read • AI Strategy
Claude Just Gave Australian Businesses a Direct Line to Real AI Adoption Data
Claude's new Economic Index connector gives Australian businesses direct access to real AI usage data, so decisions run on evidence, not vendor slides.
August 2026 • 6 min read • Technical
How to Stop Claude Code From Repeating the Same Mistakes: Verification Loops With Skills
Claude Code keeps making the same fixable mistakes. Here's how Skills turn a one-off correction into an automatic verification loop.
August 2026 • 6 min read • AI Strategy
Claude vs OpenAI Presence: Why the Access Model Matters More Than the Feature List
OpenAI's new Presence platform is enterprise-only and vendor-gated. Here's why Claude's self-configurable Managed Agents and Cowork model matters more for Australian businesses than the feature list.
August 2026 • 6 min read • AI Strategy
Why Rakuten Trusts Claude Fable 5 to Run Agents Unattended Overnight
Rakuten's Yusuke Kaji explains why Claude Fable 5 can run agents unsupervised overnight, and what the self-verification shift is worth to Australian businesses.
August 2026 • 6 min read • AI Strategy
Claude vs Google's New Gemini 3.6 Flash: What the Efficiency and Cyber Claims Actually Mean
Google's new Gemini Flash lineup leads with cheaper tokens and a security model. Here's how an Australian business should actually weigh that against Claude.
August 2026 • 6 min read • Technical
How to Secure a Software Pipeline When Claude Writes Most of the Code
Anthropic's own security team published how it defends a codebase where Claude authors 80% of merged code. Here's what Australian businesses can borrow.
August 2026 • 5 min read • Technical
What Datadog's Machine Tool Approach to Claude Code Governance Looks Like at Scale
How Datadog built a verified 'machine tool' kernel to govern Claude Code at scale, and the one-page policy Australian engineering teams can borrow from it.
Get AI automation insights in your inbox
One short email per month. What's working in Australian AI automation. Unsubscribe anytime.