https://weekly.blokhin.us/ One AI signal, one tool, one MVP idea — practical AI signals for builders, every Saturday. en Sat, 05 Sep 2026 08:00:00 +0000 https://weekly.blokhin.us/issue-013.html https://weekly.blokhin.us/issue-013.html Sat, 05 Sep 2026 08:00:00 +0000 This week's Zero to MVP Weekly: three labs shipped models built to run for hours unattended — Claude Fable 5.1, Gemini 3.8 Flash and GPT-6 Astra — and on the third day ChatGPT, Claude and Grok all reported outages inside one window, with no cause anyone will confirm. Plus what a long run needs that a short call never did, and a provider-health prober that knows about your jobs. https://weekly.blokhin.us/issue-012.html https://weekly.blokhin.us/issue-012.html Sat, 29 Aug 2026 08:00:00 +0000 This week's Zero to MVP Weekly: GLM-5.3's open weights landed on schedule — all 756 GB of them — while the MIT-licensed GLM-5.3-Flash arrived two days earlier and quietly took a third of my calls. Plus why a context window is a capacity, not a working range, and a router that prices tasks instead of tokens. https://weekly.blokhin.us/issue-011.html https://weekly.blokhin.us/issue-011.html Sat, 22 Aug 2026 08:00:00 +0000 This week's Zero to MVP Weekly: Anthropic moved computer use, Skills and the Files API to general availability, Slack gave coding agents their own channels, and GitHub put Copilot sessions in Slack and Teams. Plus the third-party verdict on Qwen3.8-27B, and the headless job agent as the MVP idea. https://weekly.blokhin.us/issue-010.html https://weekly.blokhin.us/issue-010.html Sat, 15 Aug 2026 08:00:00 +0000 This week's Zero to MVP Weekly: Alibaba shipped the first Max-class open weights ever — under a license that bills at scale — while the runnable Qwen3.8-27B arrived under plain Apache 2.0. Plus Claude Code's auto mode became the default, and a license decoder for open weights as the MVP idea. https://weekly.blokhin.us/issue-009.html https://weekly.blokhin.us/issue-009.html Sat, 08 Aug 2026 08:00:00 +0000 This week's Zero to MVP Weekly: an npm worm shipped with legitimately earned provenance and planted execution hooks in .claude and .vscode config. Plus LFM2.5-2.6B — a real agent model in 2.5 GB — and a 'who can run code on my machine?' repo auditor. https://weekly.blokhin.us/issue-008.html https://weekly.blokhin.us/issue-008.html Sat, 01 Aug 2026 08:00:00 +0000 This week's Zero to MVP Weekly: the Model Context Protocol went stateless, which drops the cost of exposing your product to agents to roughly the cost of a REST API. Plus GPT-5.6 Luna at 80% off and a confirm-before-it-bites MCP server template. https://weekly.blokhin.us/issue-007.html https://weekly.blokhin.us/issue-007.html Sat, 25 Jul 2026 08:00:00 +0000 This week's Zero to MVP Weekly: Opus 5, Gemini Flash, Microsoft's own models and a 118B open coder — four vendors repriced the frontier in five days. Plus a hands-on with Claude Opus 5 and a cost-per-task receipts MVP idea. https://weekly.blokhin.us/issue-006.html https://weekly.blokhin.us/issue-006.html Sat, 18 Jul 2026 08:00:00 +0000 This week's Zero to MVP Weekly: Kimi K3's 2.8T parameters vs Murati's Inkling — 'open weights' stopped meaning one thing. Plus a hands-on with Kimi K3 via the API and a correction-data-logger MVP idea. https://weekly.blokhin.us/issue-005.html https://weekly.blokhin.us/issue-005.html Sat, 11 Jul 2026 08:00:00 +0000 This week's Zero to MVP Weekly: OpenAI, SpaceXAI, Meta, and Tencent all ship coding models — GPT-5.6 in three tiers, Grok 4.5 at $2 — and why cost per finished task is the only price that matters now. Plus a hands-on with Grok 4.5 in Cursor and a cost-per-task-meter MVP idea. https://weekly.blokhin.us/issue-004.html https://weekly.blokhin.us/issue-004.html Sat, 04 Jul 2026 08:00:00 +0000 This week's Zero to MVP Weekly: Claude Fable 5 returns from a 19-day export-control blackout, Alibaba bans Claude Code, and China ships a free agent IDE — why your AI stack needs a fallback. Plus a hands-on with Claude Sonnet 5 and a model-continuity-kit MVP idea. https://weekly.blokhin.us/issue-003.html https://weekly.blokhin.us/issue-003.html Sat, 27 Jun 2026 08:00:00 +0000 This week's Zero to MVP Weekly: GPT-5.6 lands, but the real signal is the industry unwinding 'tokenmaxxing' — route, don't switch. Plus a hands-on with Google AI Studio's deploy-free Starter Tier and a token-spend co-pilot MVP idea. https://weekly.blokhin.us/issue-002.html https://weekly.blokhin.us/issue-002.html Sat, 20 Jun 2026 08:00:00 +0000 This week's Zero to MVP Weekly: an open model (GLM-5.2) beats GPT-5.5 on coding at a fraction of the cost, a hands-on GLM-5.2 vs Kimi K2.7 verdict, and a codemod-as-a-service MVP idea. https://weekly.blokhin.us/issue-001.html https://weekly.blokhin.us/issue-001.html Sun, 14 Jun 2026 08:00:00 +0000 A sample issue of Zero to MVP Weekly: Apple's free on-device AI for small apps, a hands-on take on MiniMax M3, and a pay-once MVP idea you could build this weekend.