Ian Cooper

Principal Engineer in London, speaker, tabletop gamer, geek. Tattooed, pierced, and bearded. The 'guv' on @BrighterCommand. Also @[email protected] He/him.

[Ian Cooper @ICooper]

I will be teaching my two-day course, Practical Messaging, at NDC Oslo this year. I suspect it will be the only public presentation this year.

https://t.co/CEKiVGLf0z

[Ian Cooper @ICooper]

@James_M_South @SixLabors Good luck. Hope that this turns into a great new chapter for you.

[Maurizio @themgmtconsult]

Been a management consultant for 20 years.

Made Partner in my 30s.
Led teams of 100+ people.
Run 9-figure client portfolios.
Lived and worked in 4 continents.

Typically, corporate IT investment would follow a common script.

Capital spent on software means a shrinking payroll.

As boards map out their strategies for the coming quarters, they are operating under the comfortable assumption that this way of thinking still holds true for AI.

But I think a fiscal reckoning is brewing there, because within the next few quarters, the current prevailing narrative of AI as a headcount killer (which we all know is vastly exaggerated) will give way to a far more punishing reality.

Instead of a clean capital-for-labor swap, executives are about to watch their IT infrastructure costs and their personnel expenses balloon simultaneously 🚀🚀🚀

It may not be fun.

First, this whole idea that generative AI can operate autonomously will shatter as early deployments attempt to scale.

Because LLMs remain inherently prone to hallucination and error, companies cannot simply fire the analysts; they will be forced to retain them (or hire new talent) to serve as high-vigilance editors.

Furthermore, because AI makes it effortless to generate code, reports, marketing collateral, etc etc organizations will soon find themselves drowning in internal output. Managing, auditing, and securing this massive influx of AI-generated material will require an unprecedented wave of human oversight....

This will ultimately EXPAND corporate bureaucracy rather than trimming it (remember the 'Scaled Agile' saga??).

Even in scenarios where entry-level automation does succeed, the math of headcount reduction will fail to balance out on the ledger.

In the coming quarters, the wage differential of the AI era will trigger *severe* skill inflation.

Replacing 5 mid/entry-level programmers does not result in a net savings of 5 salaries. Instead, it requires hiring a premium-tier AI architect whose single salary frequently eclipses the combined wages of the workers they replaced (plus tokens cost).

Companies will trade high-volume/low-cost labor for scarce/ultra-premium talent, driving TCO UPWARD despite a leaner organizational chart on paper.

Jevons' Paradox again...

AI slashes the time and cost required to draft a legal brief, design a graphic, build a software feature, and therefore executive appetite for those outputs will skyrocket.

Management will demand 10x the volume of data analysis or continuous product iterations. Because the corporate demand for output will scale far faster than the technology's efficiency gains, departments will find themselves forced to expand their human teams just to handle the sheer velocity of these new AI-driven initiatives.

Until AI achieves absolute, unmonitored autonomy (if ever), it will function not as a replacement for human labor, but as a hyper-amplifier of it.

If ungoverned, the corporate balance sheets will show that the AI boom made running the business vastly more expensive.

[Ian Cooper @ICooper]

@ThePrimeagen My gut is that you like world-building from that list, so as well as the above, N.K. Jesmin's Broken Earth. If you want character-driven, Robin Hobb's Farseer is great, though some may find it a little "slow" if they like fast-paced action.

[Ian Cooper @ICooper]

@ThePrimeagen If you want a big series like WoT, you might want to try Malazan. It's incluing, not exposition, and that discovers as you to causes some folks to bounce.

Others have mentioned Joe Abercrombie. Start with the First Law.

Scott Lynch's Lies of Lock Lamora read stand-alone

[Ian Cooper @ICooper]

@CPhilpOfficial As my MP, I am disgusted by your omission of the fact that he was convicted on two counts, in an attempt to stir racial hatred.

Please resign as my MP now. This post is disgraceful.

[Cory House @housecor]

I run 1-2 agents at once. Rarely more.

Why? Because code generation isn’t the bottleneck.

Great read.

[Ian Cooper @ICooper]

@vmpaulino Agreed. Actually, it may even be more related to the panicky mid-level, who understands patterns to deliver but is paralyzed now that they know the risks better, because they lack the judgment of likelihood and impact.

[Ian Cooper @ICooper]

As an experiment, I ran a switch from Newtonsoft to System.Text.Json in our SDD process, using Opus.

It overthinks the problem, like an anxious dev who has to ponder every possible issue.

The result is that it drowns itself in complexity and high token costs.

[Ian Cooper @ICooper]

Still, it remained an interesting experiment to me. Give the agent a trivial task, and figure out the extent to which it complicates it.

[Ian Cooper @ICooper]

If humans don't use skill and judgment when applying AI, you may overcomplicate trivial tasks. As costs rise, much of the work may not earn its cost in tokens. It's cheaper to use experience (use: lean on the compiler) than to throw the problem at an agent without thought.

[Ian Cooper @ICooper]

@GergelyOrosz A probably apocryphal story goes: IBM introduces a bug bounty for QAs to improve quality. Discovers that devs are putting bugs in the code, telling QAs where they are, and splitting the bounty.

Outcomes, not metrics.

[David Fowler @davidfowl]

Sneak peak of what's coming with @aspiredotdev 13.4:

aspire logs and aspire otel just got a major upgrade. Server-side search across logs and traces. No more client-side grep over a wall of output. Massively token-efficient for coding agents.

#aspiredev

[Ian Cooper @ICooper]

Commands are more useful than skills IMO, because generally my use case is that I want my agent to invoke something, not perhaps get it to use it if it feels its training set is not better.

[Ian Cooper @ICooper]

@captainsafia Move to London, we could get you 24 cans for that, and there are a lot of good AI employers.

[Ian Cooper @ICooper]

@marlene_zw @github @firstdrafthell He's on Bluesky if you need to contact him

[Ian Cooper @ICooper]

@marlene_zw @github You should talk to @firstdrafthell as he made Cake and Spectre.Console

[Seb Johnson @SebJohnsonUK]

An INSANELY cool UK AI startup founded by an ex-DeepMinder has just announced a $50m Series B.

This is the coolest, most under-the-radar UK AI company I've come across.

It designs, engineers and manufacturers physical infrastructure, using AI to accelerate how new technologies are discovered and brought to market.

Its first use case is AI data centres where it has already had a breakthrough with cooling fluid and signed a multiple year deal with AWS.

The team behind it are absolutely CRACKED:

> Jonathan Godwin (@jgodwin_ai): AI researcher who spent five years at DeepMind working on AI for science, engineering and advanced materials design

> James Gin-Pollock (@gin_james): A repeat AI founder who previously sold a company to Shutterstock

> Daniel Miodovnik (@Dmiodovnik): Worked across finance, government AI and advisory work to the UN

Today the company - @Orbital_Ind, has announced a $50m Series B led by @soundboy of @pluralpl.

VERY COOL

[Ronan Berder @hunvreus]

I've been complaining about AI-bros and their "20 agents working overnight" for a couple of weeks now.

That's getting old. So I thought I'd do the more productive thing and propose what I think is the Right Way™ to develop software with AI.

And I did that by stealing from the iconic agile manifesto.

https://t.co/Ktf1eZfp8T

[Ian Cooper @ICooper]

Fixing the debt, adding the quality that was previously dropped, in order to hit aggressive deadlines. AI makes it possible to hit a quality bar that was previously out of reach.

[Karri Saarinen @karrisaarinen]

We keep hearing about 10x or 100x productivity gains in engineering and knowledge work.

But outside the model labs, I haven’t seen the corresponding 10-100x revenue growth across the market or increase in quality.

So where is the productivity going?