This is 418 - I'm A Teapot, a newsletter about #Engineering #Digital #Ideas and now let's DIVE IN!
Boundaries of AI as defined by inference and verification limits
ENGINEERING - Great piece to help you understand where AI might eat the world - and what it might choke on. How task characteristics, volume of inference and ease of verification define the current limits of productive AI application:
Stay within the limits and see magic happen. Such as a bunch of maverick public servants re-building legacy government systems, at 5% the cost their systems integrators would have quoted em:
Impressive. Until you go over a cliff. Such as these note taking bots for health care professionals, that routinely get basic facts wrong. How pushing a task beyond the limits of inference and verification, ends in disaster:
DIGITAL - Bold bets all around. Over here is .txt betting on context and coherence being AIs (and teams) ultimate weakness, necessitating a whole new process and product for extracting it:
And over here, we got Block, betting that context and coherence are AIs ultimate strength and integrating factor, making management obsolete as we know it:
IDEAS - Fun times. Ready for more fun? Explore and experience AIs limits by drawing things humans can guess and AI can't. Play: Deviation Game
Does this read: I am a teapot or read this AI?
Explore more. With a font readable by humans and und readable by AI: Write: Decoy Font
Enjoy your tea.
\\[ ]]
I'm A What? 418 – I'm A Teapot is a weekly newsletter from and about the world wide web. It revolves broadly around engineering digital ideas, how, what and why to build for the web. Feel free to share, comment and contribute!
This at least is TechCrunch's take on the hype surrounding AI: Managers being delusional about AIs capabilities "and [...] uniquely prone to AI psychosis because they're sufficiently distant from the last mile of work that still has to happen to generate most value with AI."
This is both testament to a broader "AI backslash" and a welcome and productive sobering up of an almost cult like public conversation.
This is 418 - I'm A Teapot, a newsletter about #Engineering #Digital #Ideas, and today we will all get sceptical of AI together ;
IDEAS - Which is no easy feat, let me tell you that. It is just such a vexing technology, shiny, glittering, amazing and at the same time... Record numbers of Americans are turning against AI. While user numbers are steadily climbing 30%+ last year to 50%+ this year, only a meagre 16% are convinced this technology will change the world for the better.
People are amazed and overwhelmed at the same time. And so is Mat Honan of MIT Review likening it to humanities biggest crisis of the century (so far): Corona - "That March, everything was happening quickly, but it hadn't quite happened yet. We knew the world was about to change, but we did not know how. We lacked the tools to effectively measure what was going on all around us."
Once you looked at the hype from the other side, you wonder: Maybe people that wish for AI to replace all their workers are just bad at managing people?
What do you love about the Internet? Love, not like? I do love it's capability to almost capture the infinity of human creativity, passion and know how.
No niché too small, no idea too obscure, nothing too unimportant. Somewhere somebody will somehow have posted, published or otherwise shared it. Not in the walled gardens that crowd your phone, but on the myriads and myriads of private pages, boards and blogs sitting next to them.
The Internets Schrebergarten. Full of sub-, pop-, counter-, diy-, fan- and plain culture.
This is 418 - I'm A Teapot, a newsletter about #Engineering #Digital #Ideas and now let's DIVE IN!
Jane Street Trade Traffic Patterns
ENGINEERING - First stop, a dream of greatness. You are not Netflix, Google Jane Street they say. Your infra does not need micro services, lambdas, control planes and all the other shiny things of engineers wet dreams they tell us. But occasionally it's okay to dream a bit:
Second stop: A work of art! Beautiful deep dive into making 3d renders look good on the web. Even if you are 0% technical, give it a go, it's expertly explained and has plenty of interactive illustrations:
DIGITAL - Third stop, the piece that inspired this edition. A love letter and manifesto to the small Internet(s) of yonder. While the headline cries despair, the piece itself is full of hope and inspiration for us to keep the open Internet going as a space to bump into each other, to recognize us and others and to remember:
I'm A What? 418 – I'm A Teapot is a weekly newsletter from and about the world wide web. It revolves broadly around engineering digital ideas, how, what and why to build for the web. Feel free to share, comment and contribute!
---
BONUS: Top image is inspired by "Into The Spider verse.", research took me down a beautiful Internet rabbit hole. Feel free to dig in as well.
Landscape with the Fall of Icarus, Pieter Bruegel the Elder X Gemini AI
And it is almost entirely about AI. Written by humans. Not only fans, but critical thoughts, angles and ideas. From me. From others. Good words. Best words. Beast words.
This is 418 - I'm A Teapot, a newsletter about #Engineering #Digital #Ideas and today I would like to share with you:
ENGINEERING - The best argued essay on the risks of AI use I read so far. Replace lingo with your domain's and its insights apply way beyond software engineering. Key quote: "One reason that the atrophy of coding skills is concerning is the "paradox of supervision" ... effectively using Claude requires supervision, and supervising Claude requires the very coding skills that may atrophy from AI overuse.": Agentic Coding is a Trap.
This makes you wonder, are the superstars of the scene ascending to higher planes of knowledge and productivity, or are they caving in, when Simon Willison sees Vibe Coding and Agentic Engineering converge: Vibe Coding and Agentic Engineering are Getting Close.
NASA Astronaut drifting in space, NASA
DIGITAL - What a wild time to be alive. Once you dive in, it's easy to drift, outpaced by superhuman speed. The antidote is slowing down and taking agency back, great thought piece curtesy my colleague Sebastian Küpers: Have you experienced the AI drift?
You owe it to yourself. And your co-workers. This is not about mediocre people cosplaying 10x superstars. It is about putting a technology in its place and to good use: Let it brainstorm, let it edit, let it discover and write things you already know. Let it walk, do not let it drive: How to appear productive in times of AI.
IDEAS - Wow. That was tough. We deserve some fun, a game, about passwords: It's imaginatively called Password Game!
I failed at Roman Numerals. On mobile :P
Playtime is over! A fantastic startup did a crazy thing with AI, click he... na, just kidding, we're done with overexcited for today. Here is a calm camera turning pictures into poems: Poetry camera.
But, you read this far. You seem to like reading. Maybe even on paper. Here is a book. Or better, a review about one, 50 years old. The book, not the review. Still it appears relevant. Helpful to make sense of this technology beyond the hype: The Role of a New Machine.
Quote: "Artificial intelligence takes you away from your own trip. What you want to do is look at the wheels of the machine and if you like them, have fun." Unquote.
Now that you know, you wonder: How to make those expensive af teams faster? Well, Avery Pennarun (Tailscale) has some ideas about making them slower... to make them faster: Every layer of review makes you 10x slower - apenwarr
Now: Just look at the sky. Enjoy the limitless shades of soft blue. And then - Enjoy your tea.
\[ ]]
---
I'm A What? 418 – I'm A Teapot is a weekly newsletter from and about the world wide web. It revolves broadly around engineering digital ideas, how, what and why to build for the web. Feel free to share, comment and contribute!
This is I'm A Teapot, a weekly newsletter for techy creatives, creative techys and everybody else that enjoys some good reading and does not give an f- about categories. No tech knowledge needed, just a curious mind, pinky promise.
ENGINEERING
You know what made me fall in love with technology? The creativity. The riddles, the puzzles, the mind-benders. Which is why, this week I want to share two tech pieces that are about thinking outside the box and reinventing the rules:
Flowchart showing de-coupling of real and browser time.
Why making your browser time travel is the best way to record videos: Building a camera by lying to your browser about time and a great interactive exploration how Shazam actually works: How the heck does Shazam work?
DIGITAL
With that out of the way, let us dive into AI aigain ;)
Great post by Vitaly Friedman regarding AI giving better responses: Five ways of engineering context for AI Slight addendum from my side, the "thinking" anthromorphization is misleading our reading about things like this. AI "condenses" in the widest sense of the word, it is the high-pass filter of information processing tools.
What a great piece of work: The AI Atlas of Interactions, a must read imho for everyone serious about modelling human and AI-interaction beyond "I want to prompt X." level: The Atlas of AI Interaction Design
IDEAS
Golfing is not rowing - or why it is important to understand what game you are playing, in sport, business and life: Golfing Is Not Rowing
A map of happy people. Explore and feel joy little story by little story: The Pudding Happy Map
Four years ago I wrote a little internal newsletter, about #Engineering #Digital #Ideas as an experiment and to lift everyones spirits during Corona.
I kinda stopped after 52ish issues, not sure why, so, let's go again, this time in a more public setting ;) This weeks contents might be summarized under the label "exploration".
ENGINEERING
Come, take a guided tour of the (leaked) architecture of ClaudeCode: Claude Code Unpacked.
And then follow Twilio going all the way back from 100s of problem children to one superstar: Goodbye Microservices.
DIGITAL
Let's meet Mario Zechner who most politely asks us to "Slow the fuck down!". Great essay around the idea that move fast and break things gets an even worse paradigm than it already is, if AI makes it free, fast and (initially) pain less: Thoughts on slowing the fuck down.
And travel round the world to Craig Mod, who for the past couple of months went "software bonkers" and is the living anti-thesis to Mario, or is he? Software Bonkers.
IDEAS
We follow an (imaginary) software mechanic into the American rust-belt of the future: Warranty Void If Regenerated.
And we zone out listening to New Yorks subway trains playing Jazz: Every train, a note.
