research
AI Carb Counting Accuracy: I Tested Soba on 400 Meal Photos
AI carb counting accuracy, measured: Soba got 68.5% of meals within 10 g on 400 research photos. Where it fails, what fixes it, and how 8 LLMs compared.
research
AI carb counting accuracy, measured: Soba got 68.5% of meals within 10 g on 400 research photos. Where it fails, what fixes it, and how 8 LLMs compared.
MCP vs A2A vs ACP vs WebMCP in 2026: which layer each agent protocol connects, what the July 2026 MCP spec changed, the security gaps, and a reading order.
The EU Pay Transparency Directive deadline passed in June 2026, yet only 5 countries have transposed it. What developers can demand now, country by country.
The best AI models for coding in 2026 by tier: frontier, open-weight and budget, with real prices, benchmark caveats, and which one to actually use.
Ruff vs Flake8, Black, and Pylint in 2026: Astral's speed numbers, what it replaces, where it still loses, and a migration you can finish in an afternoon.
Long-form posts in your inbox roughly once a week — research breakdowns, tutorials, comparisons, the occasional review. No tracking pixels, no growth-hacked subject lines.
Or grab the RSS feed — same posts, no email needed.
I'm Maksim. By day I lead an engineering team at inDrive. After hours I ship side projects (PageBloom, NotesPilot, MyDevKit, startgaze) and write things up here when I learn something worth keeping.
The blog itself runs on an agentic publishing pipeline I built and rebuilt — a slow-moving experiment in how much of a writer's workflow can be automated without losing the voice. It writes, fact-checks, and refreshes; I edit, decide, and publish.