By Sam in AI Agents — 16 Feb 2026

Alarm Bells: Something big is happening

The rate of improvement in AI models, combined with the expanded capabilities, and the rapid reduction in costs, mean the possibilities today are frankly terrifying.

PLUS: AI agents are getting physical and why we all should pay attention to a viral lobster.

Drowning is quiet, movements are subtle, they rarely make any noise at all. One minute their head is bobbing above the water and then it is gone. - Angela Abraham

Afternoon All,

I've spent the last few years studying Data Science and AI/ML. I'm an unapologetic nerd, who is immersed in this space. I use AI to build products, and it feels impossible to keep up with the pace of new product releases. The rate of improvement in AI models, combined with the expanded capabilities, and the rapid reduction in costs, mean the possibilities today are frankly terrifying. And this is just the beginning. I deliberately write this newsletter for a non technical audience who aren't immersed in all things AI. If that applies to you then today we talk about 3 separate instances, when put together, add up to one major development that could change your life and career forever.

Today we explore a viral essay, a viral lobster and a new OpenAI employee next...

Today's dots:

Something big is happening right now
The viral AI agent that changes everything
Virtual agents, real world risks
Feel the rhythm, feel the rhyme with Data and AI

Something big is happening right now! And you need to understand it.

Here's the thing: Last week an essay on AI went viral on X. Once you see through the sales pitch/advertorial bias, the consequences of the message are monumental. The effects of AI are here, and "not in an eventually we should talk about this way. [But] in a this is happening right now and I need you to understand it" way. Matt Shumer's essay describes recent developments like "the moment you realise the water has been rising around you and is now at your chest."

Let's unpack that:

After OpenAI's GPT 5.3 Codex release, AI models now have the capabilities of doing most office/knowledge based tasks. Users have described it as having something we thought AI would never have, judgement or taste. "or something close enough that the distinction is starting not to matter."

And here's why this matters to you, even if you don't work in tech...

The AI labs made a deliberate choice. Making AI great at coding was the strategy that unlocks everything else. If AI can write great code, it can help build the next version of itself. Recursive self-improvement. AI can now write code better than most software engineers, and they're now moving on to everyone else.
Law, finance, medicine, accounting, consulting, writing, design, analysis, customer service. Some say it'll take 10 years, some say it's more like 1 - 5. In many areas it's already happening today. If you have only heard of ChatGPT and haven't tried it since 2023 because "this makes stuff up" or "this isn't that impressive", you were right back then. You're wrong now. The AI world has quickly moved on.
In 2022, AI couldn't do basic maths reliably. By 2023, it could pass the bar exam. By 2024, it could write working software and by late 2025, OpenAI said they had handed over most of their coding work to AI. Dario Amodei, CEO at Anthropic, has said that AI models "substantially smarter than almost all humans at almost all tasks" are on track for 2026 or 2027.
Let that really sit with you. They may have a multi-billion pound vested interest in pushing this narrative. But...what if they're right? If AI is smarter than most PhDs, do you really think in the right hands it can't do most office jobs? Think about what that means for your profession and your job.

If you remember nothing else: AI isn't replacing one specific skill. It's a general substitute for ALL cognitive work. AI can already read contracts, build financial models, and write marketing copy to a standard that is indistinguishable from humans. It's time for you to seriously embrace AI as more than a Google replacement. It's still early, you have time. But that window is closing faster than at any time in human history. Learn what AI can do for you before that window closes.

OpenClaw. The viral AI agent that changes everything!

Here's the thing: I don't think it's hyperbole to say that Peter Steinberger founder of OpenClaw joining OpenAI could change the course of human history. Who is Peter Steinberger, WTF is OpenClaw, and why should you sit up and take notice?

Let's unpack that:

OpenClaw is an autonomous AI agent, marketed as “the AI that actually does things.” Users have documented OpenClaw performing real-world tasks, such as web browsing, summarising PDFs, scheduling calendar entries, agentic shopping, and managing emails on a user’s behalf.
Launched less than 2 months ago, 145,000 GitHub stars and 20,000 forks and counting, suggest this AI agent may just be the breakthrough that changes everything, forever.
Kaoutar El Maghraoui, an IBM research scientist, said OpenClaw demonstrates that the real-world utility of AI agents is “not limited to large enterprises” and can be “incredibly powerful” when given full system access.
OpenClaw is open source. Meaning, anyone can freely inspect and modify its code, build new app integrations, and only pay for the costs of running the underlying AI model. Open source models are now genuine rivals to ChatGPT and Claude, and you can access them for free on any machine.
Think about what this could mean for you and the company you work for. The power of PhD level intelligence, working 24/7, for free. Now connect this story with the first one you've just read. That is why Sam Altman has hired Peter Steinberger.

If you remember nothing else: We’re getting closer and closer to everyone in the world having their own personal AI assistant. Not only that, we're getting closer and closer to the day where companies can reliably replace teams of employees with one swarm of autonomous AI agents and not miss a beat. The day OpenClaw launched is the day human history changed forever.

Physical AI. The great challenge of the next 10 years

Here's the thing: Over the past few years we've watched AI systems evolve from machine learning into an agentic operational force. Now AI is moving from language models to physical intelligence. The moment intelligence stops simply recommending or executing and instead steps into the world, learning from its own experience, everything changes.

Let's unpack that:

Previously, the truly difficult part for physical AI systems was everything humans do without conscious thought: perception, balance, navigation, adaptation to unpredictable reality. Now, whilst the physical world, is still governed by immutable laws, it has become increasingly observable. This brings new opportunities, but also new risks.
In the digital world words can be rewritten. AI recommendations can be ignored. Errors are cheap and reversible. In the physical world, mistakes aren’t just incorrect outputs, they are costly events. A manufacturing defect can be deployed at scale, at huge cost.
Real world AI removes the separation between thinking and doing. Sensors capture real-time signals, models interpret them under physical constraints, control systems translate decisions into motion, and outcomes are continuously learned from and incorporated into future behaviour.
Physical AI provides the ability to sense and act. Agentic AI provides the logic that governs when to act, how to sequence actions, and when to escalate decisions. The applications for warehouse operations, manufacturing and supply chains are enormous.
Physical AI trained on long histories of real-world behaviour accumulates depth that cannot be replicated quickly. These data moats reflect knowledge of how systems behave under stress, variability, and edge conditions.

If you remember nothing else: The defining challenge of the next decade won’t be building intelligence, that's already done. What's next is integrating this intelligence into physical, economic, and regulatory systems in ways that are resilient, accountable, and aligned with long-term objectives. The people and companies that do this best will rule the world.

AI, data, fancy footwear and the quest for Olympic Gold.

Here's the thing: I know I'm not alone in this, but every winter olympics I suddenly become an expert critic on sports I haven't seen before or at least haven't watched since the last Olympics. Bobsledding is no different. But now, thanks to the power of data and AI, I and all data and AI nerds have extra reasons to be interested.

Let's unpack that:

Prior to this year’s Games, the US Bobsleigh and Skeleton Federation partnered with a company called Snowflake to leverage the company’s AI tools to analyse data and make changes to bobbers’ movements on the track. Snowflake’s AI allowed coaches to identify which pairings worked best for two- and four-person crews.
Germany’s bobsled and luge teams are using Data Coach, their new system for analysing data from sensors installed on the sled. The team can study the best trajectories with which to approach curves and facilitate race strategy.
Not only that, BMW realised that innovations in the bobbers’ footwear could improve their performance during the crucial first phase of the race. BMW made custom spike plates that can be integrated into any type of shoe. Using 3D printing, the company was able to adapt the stiffness and shape of the studs for different shoes and needs.

If you remember nothing else: If AI and Data can used in bobsledding then it can be used anywhere. Feel the rhythm, feel the rhyme...ok that was going to be a really cheesy line so I think I'll stop writing now.