Blog

Practical guides for teams shipping production AI: agents, automation, rescuing AI-built prototypes, and web and mobile product engineering. Each article answers one buyer question with trade-offs, not hype.

Studio illustration for custom AI agents.

Gemini 4 Hydrogen and Orchestrator: When Multi-Model Agents Pay Off

Hydrogen is Google’s fast Gemini 4 workhorse for agentic coding. Orchestrator routes models inside a session. Here is when that architecture helps a product team and when it adds cost.

Studio illustration for custom AI agents.

Gemini 4 Neon for Production Agents: What to Test Before You Switch

Google’s Gemini 4 Neon targets fast end-to-end software engineering and agent workflows. Here is how founders and product teams should evaluate it for production agents, cost, and risk.

Studio illustration for custom AI agents.

Gemini Tasks: Confirmation Patterns Production Agents Should Copy

Google’s Gemini Tasks can call, pay bills, and finish chores with confirmations and isolated runtimes. Those product patterns matter more for founders than the consumer feature list.

Original ImadDhin diagram connecting human browsing, assisted discovery, and authorized agent actions to one business system.

The agentic Internet: from websites to an ecosystem of agents and bots

Cloudflare’s bot report is a starting point for a bigger business question: how should companies connect human discovery, AI assistants, and authorized agent actions?

NVIDIA Vera Rubin DSX POD announced at GTC Berlin 2026 — modular AI factory reference architecture

From LLMs to secure autonomous agents: building the next wave of enterprise and Sovereign AI

GTC Berlin 2026 reframed software around AI factories, Sovereign AI, agent harnesses, and Physical AI. Here is the stack that matters for production Super Agents, and how to ship secure custom agents on your data.

Cloudflare Workers AI decision models in October 2026: Clef 27B at $0.240, Clef-flash 9B at $0.038, Clef-omni 30B MoE at $0.150 per million input tokens

Cloudflare's new Workers AI decision models, October 2026

Cloudflare shipped Clef and Clef-flash on 1 October 2026, then Clef-omni on 9 October. They score an allowed answer instead of writing free text. Here is the family, the price, and where it fits a production agent.

Studio illustration for custom AI agents.

Agent Production: a custom AI agent that fits your business

Agent Production is the next offer for anyone who wants a custom AI agent in their business or their own case. One job, your systems, your industry rules, and a studio that stays after launch.

eTROC homepage with the AI agent answering a buyer request next to live listings

eTROC is live — an AI agent companion to buy, sell and swap

eTROC, our AI agent marketplace, is now live in Morocco. Ask the agent on WhatsApp, Telegram, iMessage, email or the web, and it finds live listings near you, compares them and drafts every message for your approval.

Studio illustration for custom AI agents.

eTROC: an AI agent marketplace companion to buy and sell

How we built eTROC, an AI agent companion that helps people buy and sell in Morocco from WhatsApp, Telegram, iMessage, email or the web, grounded in live listings near them.

Studio illustration for custom AI agents.

Agent memory and RAG development: what to store, what to forget

Most agent memory problems are storage decisions made by default. Here is how to decide what an agent should remember, where each kind of memory belongs, and when retrieval is not needed at all.

Studio illustration for custom AI agents.

How much does an AI agent cost to build and run?

AI agent cost splits into a one-time build and a recurring run bill. Here is what drives each, illustrative ranges with their assumptions stated, and how to keep cost per task visible.

Studio illustration for custom AI agents.

AI agent evaluation before production: a practical evaluation harness

An agent that looked good in five demo conversations can still fail on the sixth real one. A small, repeatable evaluation harness turns quality from an impression into a report you can rerun on every change.