Blog
Only approved posts appear here. Drafts live in Редакция.
Нет одобренных постов. Открой Редакцию.
- Bambu Lab R1: when the garage furniture-shop fantasy starts to feel real Garage furniture-shop fantasy vs panel micro-shop reality — R1 laser + conveyor, Class 4, joinery; official Bambu trailer embed; no listen-audio.
- Plugin4Shell: when SHA pinning fails on Claude Code, Codex, Copilot, and Gemini CLI SHA-pin bypass / zero-click RCE on Claude Code, Codex, Copilot, Gemini CLI — patch gates, Git-host inventory, Copilot auto-update risk. Rankings through Sep 22.
- OpenRouter Batch API: half-price async for eval queues (not live phone turns) ~50% off async batches; 70+ models; median 7 min / p90 ~1h (230k+ beta); overnight evals, embeddings, ticket backlog — not live phone turns. Rankings through Sep 21.
- Model eval as a Slack message: Descript’s Underlord loop on OpenRouter Descript Underlord: before 3 direct providers + DIY fallback; now Slack “run evals on OpenRouter” → 1–2h harness → PR. Model as parameter; Sep 20 rankings bump.
- OpenRouter shell + Files API: give any model a sandbox without owning the blast radius openrouter:shell + Files API beta: hosted Linux for any model, default-deny network_policy, promote artifacts; flash loops + shell for work; rankings through Sep 20 (DeepSeek Flash still #1).
- OpenMontage + HyperFrames: when agents should write HTML, not fight Remotion HyperFrames as HTML/CSS/GSAP for agents; OpenMontage locks remotion vs hyperframes at proposal; OpenRouter vision for capture captioning.
- Muse Secure VM: what “agent never sees the password” means for production phone agents Secure VM, Sentinel egress, authd surrogates — what “agent never sees the password” means for CRM/calendar blast radius on phone and ops agents.
- OpenRouter rankings in production: flash defaults, coding-agent traffic, when to escalate Coding/agentic apps dominate OpenRouter app traffic; weekly through Oct 1 Space Bunny Alpha #1 (~30.9T), DeepSeek V4.1 Flash #2 (~23.5T); escalate on ambiguity, irreversible actions, tool failure, and handoffs — not by default.
- OpenAI Agents API + GPT-Live: what changes for production phone agents Full-duplex voice and Agents handoffs against latency, transfers, and naturalness — the same bars from the phone-stack teardown.
- I'm ripping out my AI phone stack — here's what actually failed. Inbound agents on business phone lines, Telnyx + Vapi pain points, what I'm evaluating next, and lessons that survived the teardown.
- Cursor Projects for agentic ops Scoped projects as boundaries for agentic work — less chat forever, more shippable ops.