Blog
Field Notes For Practical DX
Short reads on AI automation, paid PoCs, B2B delivery, and fixed-scope sprint design.
Featured note
September 22, 2026
Claude Opus 5.5: Lower Prices, A Migration Worth Testing
Opus 5.5 cuts standard token rates and changes the API contract. A worked cost example and a practical checklist for deciding whether to upgrade.
September 20, 2026
How to Tell Zero Activity from Missing Dashboard Data
A practical acceptance exercise for dashboards: verify freshness, metric meaning and what happens when the feed stops.
September 20, 2026
How to Check Where a Website Enquiry Actually Goes
Trace a synthetic enquiry from the form to an owned next action before accepting a lead-capture integration.
September 20, 2026
What to Ask for in a Software Backup Restore Demonstration
A buyer worksheet for proving that saved data can become a usable application, with clear recovery limits.
September 12, 2026
When an AI Workflow Fails, What Is Safe to Retry?
A practical recovery checklist from our own media workflow: preserve finished work, verify uncertain actions and make the failed step visible.
September 5, 2026
GPT-6 Astra: Give It a Mission Worth Finishing
What changes with GPT-6 Astra, what the API costs, and a practical mission to test it: messy inputs, a mid-task correction and a deliverable you can inspect.
September 5, 2026
Microduck: The $399 Robot That Learns New Tricks
A tiny duck on roller skates, a real robot control stack, and a simulator you can try. Meet Microduck through official footage and the code behind its moves.
September 1, 2026
Claude Fable 5.1 and Mythos 5.1: What Changed
Fable 5.1 improves long-running agents, research, and office work while cutting cache costs. Mythos 5.1 adds trusted access. A practical migration guide.
August 24, 2026
Xiaomi AI Cube: What Its Local AI Prototype Actually Proves
Xiaomi showed a desktop AI prototype with its own silicon, up to 160 GB of unified memory, and local 120B plus 3B models. Here is what is confirmed, what is missing, and why it matters.
August 21, 2026
Uber's Agentic Pods: A 10-Day Playbook for AI Beyond Engineering
Uber paired its most AI-fluent engineers with domain experts and gave each pair ten days to observe, build, and ship. Sixteen pods later, the results say more about workflow design than about models.
August 19, 2026
OpenAI and Claude API Cost Calculator: Include Prompt Caching
Use our free OpenAI and Claude token cost calculator to estimate input, cache write, cache read, and output costs with verified API prices.
August 13, 2026
Grok Bot: AI Teammates With Their Own Computer
SpaceXAI launched Grok Bot on 11 August 2026. It is an early-beta teammate with its own computer, not another chatbot. Here is what it costs, who can use it, and what it does not replace.
August 13, 2026
Grok 4.6 Is Out: Agents, Pricing, And What To Do With It
SpaceXAI released Grok 4.6 on 12 August 2026. Here is the official price list, the 200K-token billing trap, how it compares to DeepSeek V4, and where Grok Bot fits.
August 13, 2026
DeepSeek V4 Flash And Pro Pricing: What The Numbers Actually Mean
Official DeepSeek V4 Flash and Pro API prices, a worked cost comparison against Grok 4.6 and other frontier models, and how to pick a tier before DeepSeek raises rates.
July 28, 2026
On-Premise AI As A Service: Your Models, In Your Building
Our on-premise AI service is live at ia.urbanodx.com: open models fine-tuned on your data, running on hardware in your building, managed end to end.
July 24, 2026
Claude Opus 5 Is Out: What Actually Changes For Your Roadmap
Anthropic shipped Claude Opus 5 on 24 July 2026, its fourth model in under two months. Same price as Opus 4.8, a 1M-token context window, and an effort dial that turns cost against capability. Here is what moves for a team with a build decision in front of it, and what does not.
July 23, 2026
Topic Filter for X: A Free Extension That Hides Topics Permanently
X only lets you snooze a topic for 24 hours. Topic Filter for X removes the topics you pick from your feed permanently, before they render, with everything running locally in your browser. Now live on the Chrome Web Store.
July 23, 2026
How Topic Filter for X Removes Tweets Before They Render
Hiding tweets with CSS leaves blank holes in X's virtualized timeline. The fix: intercept the GraphQL responses in the page's MAIN world, filter the data instead of the DOM, and back it with careful keyword matching plus an optional on-device embedding model.
July 15, 2026
GPT-5.6 Luna, Terra, And Sol: A Business Guide To OpenAI's New Models
OpenAI released the GPT-5.6 family in July 2026 with three tiers, Luna, Terra, and Sol, plus new voice models. Here is what changed, what it costs, and how to pick a tier for production workflows.
July 14, 2026
The Hidden Cost Of Subcontracting Layers In Outsourced Software
Every layer between the person who scopes your project and the person who writes the code dilutes intent, accountability, and confidentiality. Here is how to detect the layers before you sign.
July 13, 2026
How to Write Acceptance Criteria for Fixed-Scope Software
Turn a software scope into checks your team can verify, including AI evaluation, failure cases, handover and changes agreed before work starts.
July 12, 2026
Avoiding Vendor Lock-In: How To Own Your Custom Software From Day One
Lock-in is rarely a license clause. It is code you cannot access, environments you cannot rebuild, and knowledge that lives in one vendor's head. Here is a checklist to keep ownership real.
July 9, 2026
Working With a Remote, Overseas Software and AI Partner
A remote, overseas software and AI partner works when the person who builds your system is the one you talk to, you get a working demo every week, and you own everything at the end. Here is how that relationship runs, and how it differs from an offshore body shop.
July 7, 2026
How Much Does Custom AI Development Cost? A Transparent Breakdown
A custom AI build is priced by scope and risk, not by headcount-hours. Our fixed-scope packages run from £595 for a scoping teardown to £30,000-£42,000 for a full production integration, and a fixed scope protects your budget better than an open-ended hourly contract.
July 6, 2026
The B2B Partner That Builds Your Business App And Its AI Together
The partner you want is one senior team that builds the operational app and the AI inside it in the same sprint, on the same data model. Splitting the app vendor from the AI vendor just moves the risk to the seam.
June 15, 2026
Keep Working: A One-Button Prompt Deck Where Every Card Earned Its Place
Keep Working serves the next prompt your LLM session needs from a single button. The interesting part is the quality bar: the default deck only ships prompt categories that beat two baselines in blind, three-judge evaluations.
June 12, 2026
BelongHub: Turning Community Research Into A Browsable Product
BelongHub catalogs 141 verified community-building groups, games, and traditions from around the world, then turns that research into a browsable product: 11 design patterns, a quiz-driven explorer, and a starting kit for new chapters.
June 10, 2026
Claude Fable 5 And Mythos 5: What Anthropic's New Models Mean For Your Business
Anthropic released Claude Fable 5, its first publicly available Mythos-class model. What actually changed, what the safeguards and 30-day retention mean for procurement, and what to do this quarter.
June 9, 2026
Fable 5, Opus 4.8, Or Cheaper: Choosing Models For Production AI Workflows
A practical framework for deciding where Claude Fable 5 is worth $10/$50 per million tokens, where Opus 4.8 or smaller tiers win, and how to keep model upgrades a one-day decision.
June 8, 2026
How to Brief an AI PoC: A One-Page Template
A practical AI PoC brief covering the workflow, sample data, access, evaluation and final decision, with unknowns made explicit.
June 5, 2026
When The Spreadsheet Becomes The Bottleneck: From Excel To Internal App
The failure signals that tell you an operational spreadsheet has outgrown Excel, what to keep in the sheet, and how a 2-4 week sprint turns it into an internal app without losing the data.
June 2, 2026
Controlling LLM Costs In Production Workflows
Why pilot costs mislead, the cost levers ranked by leverage, and a worked support-triage example across model tiers.
May 26, 2026
GDPR-Compliant AI Workflows: Data Residency, DPAs, And Audit Trails
A practical guide to adding LLM steps to workflows that touch personal data: lawful basis, minimization, EU residency, DPAs, and audit logs.
May 21, 2026
Preserving the Heart of Japan: Protecting Culture, Temples, and Heritage
Japan's unique culture, historic temples, and social cohesion are treasures worth defending. We discuss why protecting these traditions and maintaining strict borders is vital to preserving Japan's future.
May 19, 2026
RAG Vs Fine-Tuning For Company Knowledge: A Buyer's Guide
A plain-language buyer's guide: why RAG with citations is the right default for company knowledge, when fine-tuning earns its cost, the hybrid pattern, and what a 2-week PoC should prove first.
May 16, 2026
Japan 2026 AI and DX Subsidies For Custom Software And Paid PoCs
Which 2026 Japan subsidies can fund a custom app, AI workflow, or paid PoC, and which ones only cover registered IT tools. Check order, not just the headline amount.
May 14, 2026
Japan 2026 Digitalization and AI Subsidy: Can It Cover A Custom PoC?
The 2026 Digitalization and AI Implementation Subsidy funds registered IT tools. A custom software PoC usually does not qualify unless it sits inside that registered-tool model.
May 12, 2026
AI Agents Vs AI Workflows: What Is Actually Production-Ready
Buyers hear "agents" in every pitch. A field guide to the workflow-to-agent spectrum: what reliably ships to production in 2026, what only demos well, and how to scope a first agent PoC safely.
May 9, 2026
Best AI Workflow Automation Tools For B2B Teams
n8n, Dify, Zapier, Make, and Workato are useful starting points, but the best path often turns a proven workflow into owned software.
May 8, 2026
When To Move From n8n To Custom Software
n8n is great for proving a workflow. Move to custom software when the workflow needs ownership, UX, tests, permissions, and operational reliability.
May 7, 2026
What An AI Workflow Sprint Should Prove For Japan B2B
A focused AI workflow sprint should prove one workflow, one data path, human review, and API handoff before bigger DX spend.
May 6, 2026
From Workflow Automation Prototype To Production App
The professional path is not no-code versus code. It is prototype the workflow quickly, then move critical logic into owned software.
May 4, 2026
From AI Workflow Tool Sprawl To Owned Platform
Workflow sprawl happens when every team builds automations differently. The fix is to turn the repeated patterns into owned software and APIs.
May 2, 2026
Dify, n8n, Or Custom AI Workflow: How To Choose
Dify and n8n can be excellent AI workflow starters, but custom software wins when UX, ownership, security, or product depth matter.
April 29, 2026
Buy Houses Japan Success Story: AI House Search As A Product Sprint
How Buy Houses Japan turns a plain sentence into editable filters, map results and normalized Japanese listing data, built and run by Urbano DX.
April 27, 2026
Designing Natural-Language Search Without Making A Chatbot
Natural-language search works best when AI turns messy intent into visible filters, editable choices, and product actions.
April 25, 2026
What A Real Estate Web App Teaches About AI Search UX
Real estate search is a useful model for AI UX because users bring vague intent, hard constraints, and a need to compare results.
April 24, 2026
Natural-Language Search For Web Apps: Filters You Can Inspect
Natural-language search turns a typed sentence into filters and results the user can edit. How to implement it in a web app without turning the product into a chatbot.
April 21, 2026
MVP Handover Before Scale: What Buyers Should Receive
A useful MVP sprint should leave buyers with working software, clear assumptions, source handover, demo notes, and a next-step roadmap.
April 18, 2026
API-Ready Web Apps Are Easier To Scale After A PoC
A PoC web app becomes more valuable when data boundaries, API assumptions, logging, and handover are considered from the first sprint.
April 14, 2026
DX In Weeks, Not Months
Why fixed-scope AI automation sprints help teams create visible DX proof before larger transformation spend.
April 13, 2026
AI Automation Wedges For Software Teams
Practical first AI automation use cases for teams building software, apps, websites, APIs, and LLM-powered workflows.
April 12, 2026
A Checklist Before Starting A DX PoC
What to prepare before a paid AI automation PoC so the sprint can move quickly and stay focused.
April 3, 2026
How Fixed-Price DX Sprints Avoid Scope Creep
Fixed-price DX delivery works when outcomes, exclusions, acceptance criteria, and change orders are written before build starts.
March 24, 2026
Monthly Reporting Automation Without A Data Warehouse Rewrite
Monthly reporting can be a strong first DX sprint when the work is repetitive, evidence-based, and painful for managers.
March 10, 2026
Why AI Knowledge Search Needs Citations
How citation-based knowledge search can help professional services and finance teams use AI without losing source control.
February 24, 2026
Custom App, API Integration or AI Workflow: What Should You Build First?
Choose the first software project by its bottleneck, deliverable and acceptance check, with examples for internal apps, integrations and AI workflows.
February 10, 2026
Human Review Is A Feature, Not A Delay
Sensitive AI workflows should make human review visible so automation improves speed without weakening accountability.
January 27, 2026
Why Weekly Demos Matter In Fixed-Scope DX
Weekly demos keep AI automation sprints honest by forcing visible progress, fast feedback, and scope discipline.
January 12, 2026
Support Email Triage With LLM Review
An LLM-assisted support triage workflow can reduce response delays while keeping final customer communication under human control.
December 29, 2025
PDF And Document Automation For Apps And APIs
Document automation is a strong first software sprint when PDFs, forms, or attachments need to become verified data for an app or API.
December 16, 2025
Paid PoC Or Free Pilot? How To Keep DX Serious
Why a paid DX audit or PoC creates better decisions than an unpaid pilot with no owner, scope, or acceptance criteria.