Blog
Field notes, with real numbers.
What software actually costs, what breaks in production, and how to make AI and GPU workloads fast and cheap. Every number is sourced.
12 Red Flags When Hiring a Software Development Agency
Twelve red flags when hiring a software agency: unnamed teams, quotes with no discovery, IP that never transfers, no repo access. What to ask before you sign.
White-Label Development in 2026: Selling AI Work Without an AI Team
White-label AI development for agencies: who owns the client, code and liability, vendor margin benchmarks, contract terms, GDPR Art. 28, and picking a partner.
How to Evaluate an LLM Feature Before You Ship It (2026)
Define pass/fail criteria, build a test set from real inputs, grade with code first and a calibrated LLM judge second, and gate every change on it.
FP8 and NVFP4 for Image Generation: What Changes and When It Pays
FP8 makes diffusion transformers 1.2–1.9x faster with about 40% less memory from Ada GPUs on; NVFP4 adds 1.1–1.7x on Blackwell only. Costs, quality, tooling.
HTTP2Shell (CVE-2026-111497): Is Your Node or Python App Exposed?
A Brotli decoder bug hits Node.js, Bun, Python and more. Upgrade Node to 22.23.4, 24.22.1 or 26.12.1 and pip Brotli to 1.2.1, then check what your app fetches.
PHP2Shell (CVE-2026-109618/109619): What to Patch and How to Check
PHP2Shell lets attackers run code on almost any PHP web server without logging in. Upgrade to PHP 8.5.12, 8.4.27, 8.3.36 or 8.2.35, restart PHP, then verify.
How Much Does Technical Due Diligence Cost in 2026?
Tech due diligence costs about $5K–$35K for a light review, $15K–$75K for a full specialist review and $60K–$150K+ for large M&A in 2026. What drives it.
Supabase Row Level Security: 10 Mistakes AI-Built Apps Make
The 10 Supabase RLS mistakes in AI-built apps, from RLS left off to user_metadata roles and leaky views, with bad and fixed SQL and a test for each one.
Lovable vs Bolt vs Replit vs Cursor: Which Survives Production?
None of the four is production-ready by default. Replit guards live data best, Cursor's code is most portable, Lovable and Bolt fail on Supabase access rules.
How Much Does an AI Agent or LLM Feature Cost in 2026?
AI agent and LLM feature costs in 2026: build ranges by tier, current OpenAI, Anthropic and Google token prices, and a worked monthly run-cost example.
How to Cut GPU Inference Costs for Image Generation (2026 Guide)
Cut diffusion inference costs with FP8, NVFP4, torch.compile, TensorRT, distillation and caching. Current GPU prices, cost-per-image math and benchmarking.
Fixed-Price vs T&M vs Retainer: Which Contract to Choose?
Fixed-price, time and materials, or retainer? Who carries the risk in each, deposit norms, how refund guarantees work, and the questions to ask before you sign.
How Much Does It Cost to Fix a Vibe-Coded App in 2026?
Fixing a vibe-coded app in 2026 runs about $200–$5,000 for an audit, $7,500–$40,000 to stabilize, and $50,000+ for a full rescue or rebuild. Here's why.
AI-Built App Production Checklist: Lovable, Bolt, Cursor
54 checks across 9 areas to take a Lovable, Bolt or Cursor app to production in 2026: auth, RLS, secrets, payments, infra, privacy and handoff.
Software Agency Rates in 2026: What Senior Teams Actually Charge
Senior software teams charge about $100–$250/hour in the US, $85–$165 in Western Europe and $58–$95 in Eastern Europe in 2026. Models, drivers, red flags.