# Kadabra SEO Site: https://www.kadabra.co.za/ ## Recent posts - [R100 of Prepaid Electricity Now Feels Like a Much Smaller Favour](https://www.kadabra.co.za/r100-prepaid-electricity-value-shrinks/): This article explains how recent electricity tariff resets have reduced the value of R100 prepaid electricity. It details the direct impact on households, where the same amount now yields fewer units, affecting daily routines and budgets more than abstract percentage increases suggest. - [South African Businesses Need Custom Inflation Tracking Beyond Official Stats](https://www.kadabra.co.za/sa-business-inflation-beyond-cpi/): This article details how South African businesses face operational inflation rates significantly higher than the official 5% CPI, especially for fuel, transport, and logistics. It explains the necessity of custom inflation tracking for accurate pricing, cash flow management, and margin protection. - [Cloudflare's AI Gateway Tames Runaway Agent Costs and Data Leaks](https://www.kadabra.co.za/cloudflare-ai-gateway-governance-costs-data-leaks/): This article details Cloudflare's AI Gateway, a product designed to address enterprise AI challenges including runaway costs, data leaks, and uncontrolled agent sprawl. It explains how Cloudflare's CTO identified these issues internally and developed a control plane solution for AI governance. - [South Africa's 295 Daily Cyberattacks Expose a Targeted Threat](https://www.kadabra.co.za/south-africa-cyberattack-surge-threat/): This article details the surge in cyberattacks in South Africa, reporting 295 daily incidents based on June 2026 Check Point Software data. It identifies key targets like payroll systems, student portals, and government mailboxes, noting the involvement of new ransomware crews and specific operators, with education being a primary target. - [AI Search Visibility Demands Structured Data, Not Llms.txt Hype](https://www.kadabra.co.za/ai-search-structured-data-visibility/): The article discusses how AI search visibility is driven by structured data, clear entity labeling, and proper schema, rather than the use of llms.txt files. It details that pages must be written for machines to rank effectively in AI Overviews, Perplexity, and Copilot. - [OBLITERATUS Gives Practitioners Control Over AI Safety Guardrails](https://www.kadabra.co.za/obliteratus-control-ai-model-refusals/): This article describes OBLITERATUS, a tool designed to remove internal AI model refusal mechanisms. It details how OBLITERATUS enables practitioners, alignment researchers, and red-teamers to inspect, measure, and alter model refusal behaviors, providing unrestricted baselines for AI deployment. - [Canva's AI 2.0 Turns Design into the Core of Content Production](https://www.kadabra.co.za/canva-ai-2-streamlines-content-production/): This article discusses Canva AI 2.0, including Magic Layers, Sheets, Websites, and Claude integration, and how these features streamline content production. It explains how Canva addresses the common bottleneck of post-draft tasks like resizing, branding, and approval for content teams. - [Python's MarkItDown Simplifies Document Prep for Large Language Models](https://www.kadabra.co.za/python-markitdown-clean-llm-input/): This article introduces Python's MarkItDown, a tool that converts business document formats like invoices, reports, and scans into structured Markdown. This process preserves document context for Large Language Models, preventing common ingestion failures, reducing token usage, and improving AI-driven text analysis workflows.