Insights & ideas
Notes on building modern software — web performance, AI, product strategy and the craft behind great digital products.
Recruiting is the most common business use for AI because it is full of repetitive, high-volume work. Here is what an AI hiring agent can and cannot do in 2026, and where to keep a human.
The average professional gets 121 emails a day and only 38% need a real reply. Here is how an AI email triage agent sorts, routes and drafts, and where to draw the line in 2026.
A customer portal can cut support load and win renewals, or become shelfware. What one really costs in 2026, when to build vs buy, and what to include.
AI voice cloning turned CEO fraud into a phone call your finance team can't tell from real. Why detection fails, and the verification process that stops it.
AI agents now source, screen and schedule candidates on their own. Where they cut time-to-hire in 2026, where they stay risky, and how to start.
The EU's Digital Omnibus rewrites cookie consent and parts of GDPR. Here is what changes for your website in 2026 and what to fix before it lands.
AI agents now source, screen and schedule candidates end to end. Here is what they actually automate in 2026, where they break, and how to adopt them without wrecking your hiring.
Google AI Mode now answers searches for a billion people and cites only a handful of sources. Here is how to be one of them in 2026, without gaming anything.
Fewer than a third of Google searches now send a click, and AI Mode ends 93% of them with none. Here is how businesses stay visible when the answer never leaves the page.
A fractional CTO gives you senior technical leadership a few days a month, without a full-time hire. Here is what the role covers, what it costs, and when it pays off.
Teams want a ChatGPT that actually knows their own documents, policies and data. Here is how a private AI assistant is built in 2026, what it costs, and how to keep it safe.
MCP went from a clever protocol to enterprise plumbing in 2026, with managed auth, Linux Foundation governance and API gateways adopting it. Here is what changed and what to do.
Agentic AI is turning accounts payable into a near touchless process. Here is what AP agents actually do, the ROI to expect, and how to roll one out safely.
With AI Overviews on most searches, schema markup is now how you get understood and cited. Here are the schema types that still matter in 2026 and how to use them.
79% of companies have built an AI agent; only 11% run one in production. The operating model, ownership and guardrails that get an agent past the pilot.
Low-code agent platforms ship in days but hit a wall fast. When Copilot Studio and Agentforce are enough, and when a custom AI agent is the cheaper call.
Your numbers live in five different tools and none of them agree. Here is how a custom dashboard turns scattered data into one source of truth SMEs can act on.
A real breakdown of custom software development costs in 2026, from a small internal tool to a full platform, what actually moves the price, and how to avoid overpaying.
Small language models now run on the phone. When on-device AI beats a cloud API for your mobile app, what it costs to build, and the tradeoffs in privacy, latency and battery.
AP automation, bank reconciliation and a faster month-end close are the first AI agent wins in finance. What works in 2026 and how to start.
The median company runs 25+ SaaS subscriptions and wastes a quarter of the spend. How to audit the stack, consolidate, and build what matters.
Autonomous agents run code, call tools and touch real systems. Sandboxing keeps a bad step contained. A practical 2026 guide to isolating AI agents.
Not every AI call needs a cloud round-trip. Edge AI runs inference on-device for lower latency, lower cost and data that never leaves the building.
An AI agent is only as useful as the systems it can reach. Here is how to connect agents to your CRM, ERP and databases in 2026, and when to build the integration layer versus buy it.
Most businesses know AI agents can help but not where to begin. Here is a practical way to pick your first agent workflow in 2026, one that pays for itself before you scale.
Picking one AI model and standardising on it is a 2026 mistake. How to route work between Claude, GPT-5 and Gemini 3 by cost, speed and accuracy.
Quantum computers will break RSA and ECC, and "harvest now, decrypt later" attacks are already collecting your data. What businesses should do about it in 2026.
Most AI projects fail on readiness, not technology. Use this five-part assessment to score whether your business is actually ready to adopt AI in 2026.
Platform engineering went mainstream in 2026. Here is what an internal developer platform actually is, when a growing team needs one, and how to start.
Data centres already burn about 1.5% of the world's electricity and AI is pushing that higher. Here is how to make software carbon-aware without slowing delivery.
Spec-driven development puts the spec, not the prompt, at the center of AI coding. Here is how GitHub Spec Kit and Kiro work, and when the approach beats vibe coding.
Computer-use agents see the screen and drive the mouse and keyboard like a person. Here is what they can actually do for a business in 2026, and where they still fall over.
MCP connects your AI agents to real tools and data, and to a real attack surface. Here is how prompt injection and tool poisoning happen, and how to shut them down.
Vendors are rebranding old chatbots and scripts as AI agents. Here is how to tell a genuine agent from marketing before you sign, with questions that expose the difference.
E-invoicing is going mandatory across the EU under ViDA, and several countries start in 2026. Here is the timeline, what changes for your systems, and how to prepare.
AI agents are moving from browsing to buying. Here is what AP2, ACP and x402 mean for online businesses, and how to keep your checkout reachable in 2026.
Text-to-SQL lets your team ask business questions in plain English and get real answers from your database. Here is what works in 2026, what breaks, and how to deploy it safely.
36% of companies running AI agents have no plan to supervise them. Here is a practical governance model for autonomous agents in 2026, from approval gates to kill switches.
AI agents now authenticate, hold credentials and call APIs on their own. Here is how to give them identities, scope their access and audit what they do in 2026.
EU regulators keep ruling against Google Analytics, and consent banners quietly wreck your data. Here is how privacy-first, cookieless analytics works and how to switch.
The EUDI Wallet arrives across the EU by the end of 2026, and by 2027 many businesses must accept it. Here is what eIDAS 2 means for your product and how to prepare.
LangGraph, CrewAI, and AutoGen are the three frameworks every team compares in 2026. Here is what each is built for, where each breaks, and how to pick without a rewrite later.
RAG or fine-tuning is the wrong question in 2026. Here is how to decide where knowledge and behavior should live, what each costs, and why most teams end up using both.
An AI agent that forgets every customer cannot scale. Here is how agent memory works in 2026, the architecture behind it, and how to add it without the mess.
Running five AI models in production turns into chaos: scattered keys, runaway token bills, no failover. An AI gateway fixes it. Here is what it does.
AI browsers like ChatGPT Atlas and Perplexity Comet now read and act on sites for users. Here is how to make your website agent-ready in 2026 without a rebuild.
The EU pushed its high-risk AI rules from August 2026 to December 2027. Here is what the AI Act delay actually changes for your business, and what to do now.
Agent2Agent is the open standard that lets agents from different vendors coordinate. Here is what A2A is, how it differs from MCP, and when it matters.
How to build hundreds of high-intent landing pages with Next.js and structured data, without thin content or a manual writing team.
Passkeys are now the default for secure login. What they are, why support tickets and phishing drop, and how to add them to your app without locking anyone out.
Big LLMs are not always the answer. Why small language models cut AI costs, keep data in-house, and often beat GPT-class models on your own narrow tasks.
Most AI agent projects stall on data, not models. Here is what AI-ready data means in 2026 and a practical path to get your data foundation right.
No-code gets you to market fast, then hits a ceiling. Here are the signs you have outgrown Bubble or Webflow and how to migrate to custom software without a risky rebuild.
AI coding agents are now in production at most engineering teams. Here is what the 2026 adoption data shows and how to roll them out without breaking things.
From September 2026 the EU Cyber Resilience Act puts a security clock on software products. Here is who it covers, what it requires, and what to do now.
Vendors now price AI agents per seat, per ticket, per resolution, or per outcome. Here is how each model works in 2026 and how to pick one without overpaying.
Most AI spend is justified on vibes, not numbers. Here is a simple framework to measure AI ROI in 2026, with a baseline, a formula, and the metrics that matter.
More than half your team already uses AI and under 40% of companies have a policy. A practical AI governance playbook to manage shadow AI risk in 2026.
Industry-specific AI agents sell finished work, not software seats. Here is why vertical AI is reshaping SaaS in 2026 and how to choose or build one.
RPA and AI agents solve different problems, and picking the wrong one wastes a year and a budget. A plain guide to the difference, the real costs, and how to combine both.
Multi-agent setups are the hype of 2026, but most tasks still belong to a single agent. Here is how to tell when splitting work across agents pays off and when it just adds failure modes.
Prompt engineering gets you a demo. Context engineering makes an agent reliable in production. How to design the data, memory and tools it actually sees.
Accuracy on a test set tells you little about a multi-step agent. The metrics, traces and methods that actually predict how an AI agent behaves in production.
AI agents are starting to browse, compare and buy on their own. Here is what agentic commerce means for your store in 2026, the protocols that matter, and how to make your catalog something an agent can actually buy from.
An AI agent that works in the demo can quietly fail in production for a hundred reasons. Here is how observability for agents differs from normal monitoring in 2026, what to trace, and how to catch failures before your users do.
An AI receptionist answers calls and books appointments 24/7. The real question is whether to buy an off-the-shelf tool or build your own. Here is how to decide.
Most AI agents that break in production aren't held back by a bad model, but by edge cases nobody tested. Here is how to catch those failures first.
Prompt injection tops OWASP's agentic list and shows up in 73% of AI deployments. Here is how to put real guardrails on an agent before it acts on your systems.
Serverless wins below 50,000 daily calls; containers win above 50% duty cycle. Here is the honest serverless vs containers decision for 2026, with real numbers.
Most AI agent pilots never reach production and show no measurable ROI. Here is why they stall and how to build a business case that survives real traffic.
llms.txt is a markdown map that points AI crawlers at your best pages. Here is what it does, what it honestly does not, and how to ship one that helps.
A production AI agent can cost $10 to $100 per session. Here is how model routing, prompt caching and tighter context cut token spend by 60 to 80% without breaking the agent.
From 2026 the EU's Digital Product Passport starts to bite for batteries, textiles and electronics. Here is what the DPP actually demands from your software, and how to build for it without guessing.
AI sales agents promise to fill your pipeline on autopilot. Here is what an AI SDR actually does well in 2026, where the fully autonomous version falls apart, and how to deploy one that books real meetings.
Invoices, contracts and forms still trap most business data in PDFs and inboxes. Here is how intelligent document processing works in 2026, what it actually costs, and how to pick the first workflow that pays for itself.
A custom AI agent runs from $8K for a simple assistant to $150K+ for an autonomous one. Here is the honest breakdown of what drives the budget and where the money actually goes.
35% of teams have already replaced a SaaS tool with custom software, and 78% plan to build more. Here is how to decide between building, buying, or a low-code platform.
By 2026 roughly one in ten support calls is handled end to end by AI voice agents. Here is where they actually work, what they cost, and how to deploy one without wrecking CX.
WhatsApp is where your customers already are. Here is how the Business API works in 2026, what it costs per conversation, and how to wire it into your own software.
83% of enterprises plan to move workloads off public cloud. Here is when repatriation actually saves money, when it does not, and how to do it without a big-bang migration.
Next.js 16 makes caching opt-in, turns params into promises, and swaps middleware for proxy.ts. Here is how to upgrade an App Router app without breaking it.
A practical guide to choosing a business automation platform in 2026. How n8n, Zapier and Make compare on AI agents, pricing and data control.
EU data sovereignty is reshaping where companies host data and run AI in 2026. What the CLOUD Act, GDPR and the AI Act mean for your stack, in plain terms.
From August 2026, AI chatbots and generated content must be disclosed under Article 50 of the EU AI Act. Here is what your business actually has to do.
The Model Context Protocol is becoming the standard way to plug AI into real systems. Here is what an MCP server is, why it matters, and when to build one.
Buying SaaS looks cheaper until year three. Here is the build-vs-buy framework we use, the real total cost of ownership, and when custom software actually wins.
92% of developers now use AI to write code, but 45% of it fails OWASP checks. Here is how to ship vibe-coded software to production without the breakages.
INP is the most failed Core Web Vital in 2026, and 43% of sites still miss it. Here is how to fix interactivity, LCP and CLS so Google can actually rank you.
A real SaaS MVP runs $15K to $80K in 2026. Here is what actually drives the budget, where AI-assisted builds cut cost, and how to scope an MVP you can fund.
A headless CMS promises speed and flexibility, but it is not free. Here is when a decoupled stack pays off for a marketing site, and when it is overkill.
Portugal pairs EU talent, the London time zone and 30 to 45% lower costs. Here is how nearshore software development works and how to pick a partner who ships.
Full rewrites fail more often than they ship. How to modernize legacy software in phases, with the 7 Rs, the strangler fig pattern and AI-assisted refactoring.
Native is not the default anymore. Here is how progressive web apps compare on cost, reach and performance, and when a native build still earns its keep.
Ranking on Google no longer guarantees you appear in ChatGPT, Perplexity or AI Overviews. Here is how Generative Engine Optimization gets your brand cited in AI answers.
Cross-platform apps now cost 35 to 50% less than dual-native builds. Here is an honest breakdown of what drives a mobile app budget and how to choose React Native, Flutter or native.
Most Next.js sites leave rankings on the table with a few avoidable mistakes. Here is the App Router SEO playbook we use to ship sites that Google can actually read.
The EAA is enforced and ADA Title II lands in 2026. Here is what website accessibility law actually requires, why overlay widgets backfire, and how to build it in properly.
Most startups waste 25 to 35% of their cloud bill on idle and over-provisioned resources. Here is a practical FinOps playbook to cut spend without slowing down.
Retrieval-augmented generation lets an AI assistant answer from your verified knowledge base instead of guessing. Here is how RAG works and how to deploy it safely.
A modern website is no longer a brochure. It is your most reliable salesperson, recruiter and credibility signal. Here is why it matters for early-stage growth.
Most React performance problems come from a handful of recurring patterns. This guide covers the techniques that deliver the biggest wins with the least complexity.
AI agents can now handle real, multi-step work, from support to operations. Here is a grounded look at where they help today and how to adopt them responsibly.
Newsletter
Occasional notes on software, design and what we're building. No spam — unsubscribe anytime.
