AI.
Remiam has been shipping AI in production since 2023 — starting with Finding Mastery Bot in the earliest days of the OpenAI API, when LangChain was still pre-1.0 and 'production retrieval' was barely a phrase. The studio's approach is grounded: AI earns its place where it does something humans genuinely can't (or shouldn't, repeatedly) — drafting, summarisation, retrieval over a private corpus, image labelling, content classification. The notes catalogue documents this view in detail across multiple essays, and the work itself spans Finding Mastery's editorial chatbot, the Recruit Pipeline MVP's enrichment layer, the experimental Segment Analyzer (GPT-4 Vision over scraped retail interiors), and several client features that aren't named publicly.
Finding Mastery Bot
An AI chatbot built for Finding Mastery — the Michael Gervais podcast on the psychology of high performance — trained on their entire podcast back-catalogue so their internal editorial team could ask questions, surface guest insights, and write better content. Shipped in 2023 in the earliest days of production OpenAI, and the start of an ongoing research collaboration with Finding Mastery on applied AI.
Recruit Pipeline
A B2B SaaS MVP built end-to-end for a recruitment-tech client. Recruiters log jobs, run hiring campaigns, manage companies and candidates through custom-field workflows, let AI populate routine information, and move candidates through the pipeline — interviews, comms, PDF reports — without leaving the system.
Segment Analyzer
An experimental AI product built for a US strategy client. Type a location and an industry sector — the tool scrapes that location, gathers interior photos, runs them through GPT-4 Vision to label features and surface percentage breakdowns, then lets you compare industries head-to-head on interior-design trends. Replaced weeks of in-person fieldwork with minutes of automated analysis.
- 5 Oct 2026→
Heat is the hard part.
On 1 October Google put four of its Trillium TPUs into orbit on a Planet-built satellite, the first step of Project Suncatcher. The chips had already passed their radiation tests on the ground. The open question is the one we meet on every enclosure we build: where does the heat go.
- 4 Oct 2026→
Your Android app has a cursor now.
The first Googlebooks ship on 4 October in the US and on 5 October in the UK. They are $899 laptops running an Android-based desktop OS, which means the Android app you built for a phone or a tablet fleet is about to meet a trackpad, a keyboard and a window someone can drag to any size.
- 2 Oct 2026→
Finding the bug is the fast part now.
On 30 September Google released Gemini 4 Argon to cyber defenders first, a model it says can find, validate and patch critical vulnerabilities on its own. If finding the flaw stops being the slow step, the slow step becomes getting the fix onto every screen, kiosk and box you have already installed.
- 28 Sept 2026→
Agents read the accessibility tree.
On 24 September AudioEye published a study of 1,560 AI agent runs: on the least accessible site tested, agents finished 31% of their tasks, and on an accessible version of the same site they finished 96%. The work we used to justify for screen readers now decides whether software can use your site at all.
- 20 Sept 2026→
The pause is the product
Google has shipped a voice model that talks while it thinks. The benchmark gain is real. The more useful lesson is that the silence between a question and an answer is now something you design.
- 17 Sept 2026→
Ready to code is a spec now
Microsoft has put a memory bandwidth floor on what counts as a developer machine. The clean desktop got the headlines. The 250GB/s is the part that changes how a machine gets bought.
- 14 Sept 2026→
The forecast refreshes every hour now.
Google DeepMind's WeatherNext 3 produces a fresh global forecast every hour, with temperature and moisture at about 5 kilometres, and developers can query it today. For anything we build that runs outdoors, the interesting change is not the accuracy. It is that the weather decision no longer has to be made once, at breakfast.
- 11 Sept 2026→
Siri has a daily allowance now.
iOS 27 reaches the public on Monday, and Apple's own support page puts daily usage limits on Siri AI, the cloud image tools, and any developer app calling Apple Foundation Models in Private Cloud Compute. Free AI in the operating system now has a meter on it, and anything built on top of it has to know what happens when the meter runs out.
- 6 Sept 2026→
The input you do not ask for
IFA 2026 was full of appliances that work out what you want instead of asking. The design problem that creates is not accuracy, it is what happens when the guess is wrong.
- 2 Sept 2026→
The protocols settled. The processes did not.
On 17 August the last big agent protocol moved into the same neutral foundation as the other one. The interoperability argument is basically over. Meanwhile only about one company in nine has an agent actually running in production, and that gap is not a technology problem.
- 30 Aug 2026→
A humanoid in a shop is an installation.
XPeng Robotics raised over 900 million dollars on 24 August and said the first IRON humanoids go into its own stores and campuses. That is not a hiring decision, it is a site fit-out, and the questions that decide whether it works are the ones we ask on every install.
- 28 Aug 2026→
The newest chip is in the cheapest Mac.
Apple put its newest silicon in the 899 dollar Mac mini and shipped the 5,499 dollar Mac Studio on last year's generation. Generation and tier have come apart, which means "get the newest one" has stopped being a hardware spec.
- 19 Aug 2026→
Stripe bought the router.
Bloomberg reported on 16 August that Stripe has agreed to buy OpenRouter for more than seven billion dollars. A payments company paid a payments price for the layer that sits between an application and four hundred models. That tells you where the durable part of an AI system actually is.
- 17 Aug 2026→
Thinking is a setting now
Visual Studio 18.9 shipped on 11 August with a thinking-effort control in the model picker. How hard the model reasons is now a dial you set per task, and it bills accordingly. That makes reasoning depth a design decision, not a vibe.
- 12 Aug 2026→
The agent can live in the building now
Meta put a 30 billion parameter agentic model out under Apache 2.0 on 10 August, small enough for one consumer GPU. The licence is the news, not the benchmark table.
- 10 Aug 2026→
The browser was the wrong container
OpenAI switched off its Atlas browser on 9 August, nine months after launching it. The capability did not fail. The container did, and that is a much more useful lesson than it looks.
- 26 Jul 2026→
AMD showed a rack. We are watching the edge.
AMD's Advancing AI 2026 keynote led with a full rack of 72 GPUs for frontier AI. The announcement that actually changes what a studio can build was further down the slide deck: embedded silicon and a turnkey robotics platform small enough to sit inside an installation.
- 16 Jul 2026→
A keyboard for the agent fleet
OpenAI has released a 230 dollar keyboard for running Codex agents, with status lights, shortcut keys and a dial for how hard the model thinks. The gadget is a novelty. The instinct behind it, that a fleet of coding agents needs a control surface, is exactly right, and we already run one.
- 5 Jul 2026→
When inference is almost free
The price of running a capable model keeps falling by roughly an order of magnitude a year. That does not just make AI features cheaper, it moves the line for where you can put a model in the first place.
- 3 Jul 2026→
What an agent is allowed to touch
Microsoft has put an agent sandbox into Windows itself. The interesting part is not the feature, it is that the question it answers, what is this thing allowed to touch, is the question we have been answering by hand on every AI system we ship.
- 29 Jun 2026→
A sub-1nm chip is a roadmap, not a part.
IBM has put nearly 100 billion transistors below the 1nm line. It is a real milestone and the right kind of news, but it is a roadmap for the decade, not a part you can specify this year.
- 26 Jun 2026→
NVIDIA's RTX Spark and the case for local inference
NVIDIA put 128GB of unified memory in a laptop and called it an AI PC. The interesting part is not the petaflop, it is where inference starts to run.
- 25 Jun 2026→
The AI memory crunch is now a tax on everything you ship.
Memory got expensive in a hurry. DRAM rose around 90 percent in the first quarter of 2026 as the big makers moved their lines to AI memory. If you ship anything with RAM in it, on a device or in the cloud, that is now a number you have to design around.
- 14 Oct 2025→
AI in client systems: building with it, not around it.
Clients in 2025 don't ask us 'should we use AI?'. They ask us where it goes, what it costs to run, and what it can't do safely. Here's the answer we've been giving.