NOOR YASSERPRODUCT ENGINEER
عربيBook a call
JOURNAL / PRODUCT ENGINEERING

Engineering decisions.
Explained in practice.

Practical guides to reliable systems, from tenant isolation and integrations to AI retrieval and product operations.

Page 4 of 7 · Articles 37–48 of 84

Original conceptual illustration of an always-on agent control plane connecting task ledgers, memory, permissions, sandboxed compute, approvals and an emergency stop; not a real OpenAI interface.
AI agents · Durable supervision37

OpenAI Dots need durable agent supervision

OpenAI's always-on Dots move agents from request handling to durable operations. Here is the task, memory, permission and recovery architecture that implies.

9 min readRead article
Original conceptual illustration of a product outcome branching into user opportunities and competing solutions, then narrowing through evidence to one experiment; no real product data.
Product management · Discovery38

Prioritize product opportunities before solutions

A practical discovery workflow for moving from a measurable outcome to user opportunities, risky assumptions and the cheapest evidence that can change a decision.

7 min readRead article
Original conceptual illustration of an AI coding agent retrieving Zid documentation through a controlled context layer, passing an OAuth boundary and reaching merchant APIs; not a real Zid interface.
Saudi commerce · AI integration39

Zid AI Agent Skill and MCP: a verified integration workflow

Zid provides an AI Agent Skill and an MCP server for its developer documentation. Here is how to turn them into a safe, testable integration workflow.

8 min readRead article
Original conceptual illustration of an AI agent moving through a hosted browser, origin approval gate, sign-in handoff, recovery loop and audit trail; not a real OpenAI interface.
AI agents · Computer use40

OpenAI Agents API computer use needs a trust architecture

OpenAI added hosted browser computer use to the Agents API. The important work is designing approvals, authentication, recovery and verification around it.

7 min readRead article
Original conceptual illustration of an older amber AI runtime passing tools, context and evidence through a transparent migration gate into a violet production model, with a rollback lane; not a real interface or benchmark result.
AI · Model releases41

Claude Opus 5.5: the production migration is more than a model swap

Anthropic released Claude Opus 5.5 with lower prices, a 1M context window and API changes that can break thinking, tool-use and computer-use integrations.

7 min readRead article
Original conceptual illustration of a versioned test dataset passing through several anonymous AI candidates and a transparent evaluation gate, with only the passing candidate reaching a successful user outcome; not a real interface or benchmark result.
Product engineering · AI evaluation42

Your AI evals need a portable product contract

A Product Engineering framework for turning AI evals into portable release evidence tied to user outcomes, cost, latency and reversible rollout decisions.

8 min readRead article
Original conceptual illustration of anonymous customer segments passing through a policy router toward different payment cards and shipping parcels; not a Salla interface or evidence of a real deployment.
Saudi commerce · Checkout policy43

Treat Salla customer groups as checkout policy

A practical architecture for using Salla customer groups to control payment, shipping and offer eligibility without turning segmentation into fragile manual configuration.

7 min readRead article
Original conceptual illustration of product cards being reordered through a pointer drag path and an equivalent keyboard-focused move-button path; not a real product interface.
Product design · Accessible interaction44

Drag-and-drop needs a second path

A practical Product Design framework for reorder flows that work with pointer, touch, keyboard and assistive technology without turning accessibility into a hidden shortcut.

7 min readRead article
Original conceptual illustration of an AI model-routing core balancing capability, cached context, cost and production workflows; not a product screenshot.
AI · Model releases45

GPT-6.1 Sol: what developers should test before migrating

OpenAI released GPT-6.1 Sol with stronger coding, agent and computer-use claims at Sol pricing. Here is the practical evaluation and migration plan.

5 min readRead article
Architecture diagram showing three store queues sharing a fair scheduler before separate API budgets.
Saudi commerce · API capacity46

Salla API limits: fair queues before more workers

Design per-store API budgets and fair queues for Salla integrations, with controlled retries, uncertain-write recovery and visible synchronization progress.

4 min readRead article
Original conceptual illustration of a feature-flag gate splitting user cohorts into gradual rollout paths, with telemetry, rollback and cleanup loops; not a real product interface.
Product engineering · Progressive delivery47

Feature flags need an exit plan, not just a toggle

A product-engineering contract for feature flags: stable cohorts, real exposure events, outcome guardrails, failure defaults, progressive rollout and safe cleanup.

8 min readRead article
Original conceptual illustration of an ecommerce return splitting into physical inspection and financial refund paths that reconcile before completion; not a real Zid interface.
Saudi commerce · Returns architecture48

Zid reverse orders: separate return approval from refund execution

A practical architecture for Zid returns that separates approval, physical inspection, inventory decisions and refund execution while preventing duplicate financial actions.

8 min readRead article