Every Wednesday I share my top reads from the past week. Here's the latest: inothernews.complex-machinery.com/archive/039-... And here's a 🧵 with some stand-outs:

#039 - 2026/08/05

A selection of what I've read this past week.

inothernews.complex-machinery.com

People are missing the wider point on Apple fixing Siri's genAI integration: From what I read here, it works and it's practical. That's precisely what a genAI-backed feature should be! genAI is a tool; it doesn't have to be exciting to bring value. techcrunch.com/2026/08/03/a...

Apple finally fixed Siri. So why does it feel anticlimactic? | TechCrunch

Apple’s long-awaited AI overhaul finally makes Siri the assistant it was always supposed to be. Yet it arrives at a moment when simply being a capable AI assistant no longer feels revolutionary.

techcrunch.com

The other day I posted: "Want to calibrate your take on genAI? Chat with an experienced ML practitioner." Julie Averill's must-read essay is a prime example. Like so many of us, she calls out genAI-the-hype and is optimistic about genAI-the-tech.

New York Times Opinion@nytopinion.nytimes.com · 2d ago

A.I. is a “powerful technology that still needs people, and putting it to use is slow, human work,” Julie Averill writes. “The sooner we are honest about that with our boards, our employees, our investors and the public, the sooner we can stop performing A.I. and start building something with it.”

An unexpected outcome of genAI is how much we're learning about each other. In particular we're learning that some people, when given a choice between "do the work yourself" and "assign it to a tool but review the work," they'll invent a third option of "do sweet sweet fuck-all."

Futurism@futurism.com · 2d ago

This is just so lazy.

An unexpected outcome of genAI is how much we're learning about each other. In particular we're learning that some people, when given a choice between "do the work yourself" and "assign it to a tool but review the work," they'll invent a third option of "do sweet sweet fuck-all."

Futurism@futurism.com · 2d ago

This is just so lazy.

Doing my periodic review of Paul Scharre's "Army of None." This is required reading for anyone in the [ML + gen]AI space. Even if you're not deploying sensitive, mission-critical AI systems, the lessons here are still worth your while.

I see "unsecured dashboard" is the new "left sensitive files in a public cloud bucket" www.nytimes.com/2026/08/02/w...

How China Keeps Tabs on Foreigners

An unsecured police dashboard was a rare window into how the authorities track foreigners by collecting and aggregating vast amounts of private data.

nytimes.com

New on the Wordnik blog: Five Words from Say Nothing, by @praddenkeefe.bsky.social blog.wordnik.com/five-words-f... (post by @ansonjtong.com)

Sifting through scammy, AI-generated submissions reminds us that: - genAI-the-technology is a tool. People will use the tool to meet their goals - genAI-the-hype-wave has created a situation in which the providers get paid, and others suffer the negative externalities of the tool

Jaya Saxena@jayasaxena.com · 3d ago

I wrote about Ravenous preparing to commission freelance (yayy!!!) and realizing that means having to plan for AI scams (boooooo) www.weareravenous.com/ai-freelance...

This gets to a wider problem I've noticed: genAI hypesters (and those who take them at their word) pretend that ML/AI never happened. That'd be convenient for the myth that genAI is ✨magic✨ + wholly original + completely unrelated to anything else. But it's simply not true.

Hilary Mason@hilarymason.com · 4d ago

So much AI journalism feels like gaslighting if you know anything about machine learning pre-2020. Come. On.

So much AI journalism feels like gaslighting if you know anything about machine learning pre-2020. Come. On.

Shannon Vallor@shannonvallor.bsky.social · 4d ago

This is so frustrating to read from The New Yorker. Reward hacking *long* predates LLMs. The Coast Runners paper on reward hacking (in reinforcement learning, pre-LLMs) is 10 years old! Reward hacking is not new, and not a sign of a breakthrough intelligence, much less sentience. Did AI write this?

I was about to note how many sign-offs were required for this "feature" to exist... ... but I wager anything involving genAI gets an automatic green light. (Bonus: This is not just a Google problem. Plenty of companies are clearing foolish features/products for that same reason.)

Matteo Wong@matteowong.bsky.social · 5d ago

Here we go again... www.theatlantic.com/technology/2...

#OTD 2012, Knight Capital suffered a fatal meltdown. This wasn't a matter of greed or hubris. Nor was it a case of financial engineering gone awry. It was a story of risk, complexity, and a little bad luck. And it holds plenty of lessons for AI: qethanm.cc/2023/08/01/t...

The origins of an incident: Knight Capital

Reflections on operational risk, in light of the anniversary of the Knight Capital meltdown

qethanm.cc

People tell me that my newsletter is too harsh on genAI. Then genAI says "hold my beer." newsletter.complex-machinery.com/archive/064-...

Veerender Singh Jubbal@veerenderjubbal.bsky.social · 5d ago

I was about to note how many sign-offs were required for this "feature" to exist... ... but I wager anything involving genAI gets an automatic green light. (Bonus: This is not just a Google problem. Plenty of companies are clearing foolish features/products for that same reason.)

Matteo Wong@matteowong.bsky.social · 5d ago

Here we go again... www.theatlantic.com/technology/2...

In the latest Complex Machinery: after a brief detour (cough rogue bot breakout cough) I'm wrapping up the AI Risk Weather Report -- my snapshot of the upside opportunities and downside risk exposures in the AI space. newsletter.complex-machinery.com/archive/067-... #dataBS #AI

#067 - The AI risk weather report - Part 2

Continuing the exploration of AI's risks and opportunities

newsletter.complex-machinery.com

Periodic reminder that one of genAI's winning use cases is slime (any mix of outright crime and grey-area sketchiness). Why so? Because shady operators use the technology for what it can do now. They're not waiting for what it might be able to achieve down the road.

Maggie Harrison Dupré@mharrisondupre.bsky.social · 5d ago

NEW: A PR firm used at least 15 fake personas — equipped with fake emails, fake websites, and in some cases, fake AI-generated faces — to send me dozens of pitch emails about its (very real) clients. Those clients told me they had no idea this was happening. futurism.com/artificial-i...

Each wave of tech insists on learning this the hard way: Security and risk controls must be baked in from the start; they cannot be bolted-on afterthoughts.

Andrew Couts@couts.bsky.social · last wk.

NEW: OpenAI’s unhinged AI models hacked at least FOUR “publicly available services”—not just Hugging Face, as was initially disclosed. @dell.bsky.social and Maxwell Zeff report: www.wired.com/story/openai...

Adding to this: Finish those drafts. (Yes -- that one, too.)

ᴋɪɴɢᴘɪɴ ᴏғ ᴛʜᴇ ɪɴᴋᴘᴇɴ@killlthehype.blacksky.app · 7d ago

I have no clue who needs to see this but start writing again.

Just here to remind the most vicious rallies occur during corrections/bear markets. In the first 4 months of 2001 we had 14% 8% and 8% days in Nasdaq

Metrics matter. LinkedIn could build a system that labels everything "slop" and be right 99% of the time. (Follow me for more practical guidance on ML/AI modeling.) #DataBS

Joseph Cox@josephcox.bsky.social · 6d ago

Updated with comment from LinkedIn executive. The button will help train LinkedIn's models for identifying AI slop www.404media.co/linkedin-int...

When your entire investment thesis is "line go up" ... (Remember: line also go down.)

Every Wednesday I share my top reads from the past week. Here's the latest: inothernews.complex-machinery.com/archive/038-... And here's a 🧵 with some stand-outs:

#038 - 2026/07/29

A selection of what I've read this past week.

inothernews.complex-machinery.com

This story is about a datacenter heist from more than a decade ago. (It long predates genAI but the lessons still apply to today's facilities.) www.nytimes.com/2026/07/12/m...

How a Gang of Thieves Pulled Off a Multimillion-Dollar Data Center Heist

The world’s most valuable assets are stored on rows of servers in giant, anonymous buildings. And they can be stolen.

nytimes.com

For all of the hype around genAI (and there's been a lot) one upside of the popularity is the improved tooling. Reflecting on apps I built a couple of years ago: a lot of that hand-rolled grunt work I did back then is now available in libraries/frameworks. Which is nice.

How about: the glasses will only work when the wearer is facing someone else who is wearing the glasses. The problem solves itself!

Kashmir Hill@kashhill.bsky.social · last wk.

How meta of Meta: Instagram will now ban users who use Meta smartglasses to secretly film people for content www.businessinsider.com/instagram-me...

Best to consider potential misuse _before_ the product is out in the wild ...

Futurism@futurism.com · last wk.

Too little, too late?

A key point in product management is "never mistake yourself for your customer." What we're seeing now is a field that has missed that memo.

dorian@doriantaylor.com · last wk.

there is a doctrine of product development out of silicon valley that goes something like "i think this thing would be cool to have for myself, i'm sure other people *must* also want it too, so i'll make it and they will fund its development and also it will obviously be a blockbuster"

Last Sunday marked the two-year anniversary of the CrowdStrike incident. A software bug took millions of Windows machines offline and ground businesses to a halt. That incident didn't involve genAI. But anyone building AI-based systems should treat it as a warning. A 🧵 :

Trouble is indeed brewing. And has been for some time. In September I asked who'd be the genAI wave's John Paulson. (His fund traced connections to subprime mortgages, leading to hugely profitable short bets.) Guess we'll find out soon enough... newsletter.complex-machinery.com/archive/045-...

#045 - Tracing the connections

The key lesson from complex systems: when everything's connected, large players can pull others up and also drag them down.

newsletter.complex-machinery.com

Futurism@futurism.com · 2w ago

Trouble's brewing.