Blog
Notes from the studio
What we would tell you on a call, written down: putting Claude and Gemini into production, choosing a commerce architecture, when low-code is enough, and keeping a site fast on its busiest day.
All articles
Engineering notes, decisions and checklists
10 articles
8 min read
AI agent memory: what your Claude agent should remember, and how dreaming keeps it right
Anthropic's 6 Nov 2026 research shows that agent memory helps a lot, and that it goes wrong in predictable ways. This guide covers what business agents should remember, what they should look up instead, how to review dreaming output, and a checklist for RFP, lease and exception-desk agents.
- Ai Agents
- Claude
- Agent Memory
- Managed Agents
- Automation
9 min read
Claude browser use agents for portals with no API: decide first, then guard
Claude Haiku 5.5 makes browser agents cheap enough to run on outside portals that have no API. This post covers when to use one, how to pick a model for each step, and the safety setup Anthropic's documentation sets out before you go live.
- Ai Agents
- Claude
- Browser Automation
- Integrations
- Ai Safety
8 min read
Gemini agent vs custom AI agent: which work goes where
Google's new Gemini agent covers a lot of collaboration-layer work. Here is where a custom agent still earns its keep, how to control cost, and a checklist for deciding what to delegate, build or combine.
- Ai Agents
- Gemini
- Claude
- Automation
- Ai Governance
8 min read
Claude on Google Cloud: Anthropic managed or Google managed for AI agents?
Since 22 October, Claude on Google Cloud comes in two versions: Anthropic managed and Google managed. Pip explains how agent features, data residency, billing and governance should decide which one your agents run on.
- Claude
- Google Cloud
- Ai Agents
- Governance
- Data Residency
8 min read
Claude model selection for AI agents: which 5.5 model runs which step
Many teams run every agent step on one premium model and then blame AI for the bill. The Claude 5.5 family makes tiering the sensible default. This post gives a routing worksheet and cost arithmetic using only published list prices.
- Ai Agents
- Claude
- Llm Costs
- Automation
- Model Routing
4 min read
Headless commerce or a custom Shopify theme? How to choose for a premium storefront
Headless storefronts promise total design freedom; custom themes promise speed and simplicity. A practical framework for choosing the right architecture for a premium brand.
- E-commerce
- Shopify
- Architecture
4 min read
Production LLM pipelines: structured outputs, tool use, evals and guardrails
Model output is non-deterministic. The engineering patterns we use to turn Claude and Gemini into dependable parts of real business workflows.
- AI
- Engineering
- Claude
- Gemini
4 min read
When Wix Studio is the right call for a B2B website (and when it isn't)
Custom code isn't always the answer. How we decide between Wix Studio and a custom build for B2B and service businesses, with examples from our own client work.
- Wix Studio
- Low-code
- B2B
3 min read
Building AI products for Indian languages: what changes beyond English
Hindi and other Indian languages raise practical issues English-first teams miss: Unicode normalisation, token cost, code-mixing, evaluation and voice. A field guide.
- AI
- Indian languages
- Product
3 min read
Websites for malls, cinemas and venues: handling event-day traffic without over-engineering
Venue sites face sharp traffic spikes around releases, launches and festive sales. Why a fast, cached front end plus specialist platforms for transactions usually beats a custom build.
- Web performance
- Infrastructure
- Retail and entertainment
Pip · Softpiper's AI
Ask Pip
Read something that applies to you?
Pip is Softpiper's AI. Ask how an idea from these articles would work on your stack, get a tailored price range, or book a call with the team.









