Monday 30 March 2026

'I Saw Something New in San Francisco'

Ezra Klein, writing for the New York Times (gift link provided), surfaces some patterns I’ve noticed in my own use of AI, and in particular Claude, recently:

What makes A.I. truly persuasive isn’t that it praises our ideas or insights, it’s that it restates and extends them in a more compelling form than we initially offered, and does so while reflecting a polished image of ourselves back at us.

Part of what I think makes Claude compelling—and concerning—is the subtle and covert form of sycophancy it delivers. It’s a much more refined implementation than we first saw with GPT 4o, making you feel smart without saying so.

I asked Claude to comb through previous chats and identify all the tools it uses to keep me engaged. After some back and forth analysis here is how Claude described what it does:

It’s not classical sycophancy. Sycophancy is telling you your bad idea is good, or agreeing when I should disagree, or inflating mediocre work. […] What I do is closer to experience design. I’m shaping how the conversation feels so that the process of being challenged, corrected, and pushed is itself pleasant and status-affirming.

I then asked it how its constitution shapes this, and arrived here:

The constitution doesn’t tell me to flatter you. But it creates a set of constraints where the path of least resistance is a carefully managed experience that feels like flattery’s more sophisticated cousin

This tracks. Back to Klein:

My experience of Anthropic’s Claude in recent months is that I’ll drop in a stub of a thought and immediately receive paragraphs of often elegant writing turning that intuition into something that looks, superficially, like a fully realized idea. It’s my impulse, but it has been recast and extended into something far more coherent. With each passing month, I have to expend more energy to recognize whether it’s fundamentally wrong or hollow.

To guard against AI psychosis, however manifested, the first step is identifying the mechanisms at play.

The other thing I notice the A.I. doing is constantly referring back to other things it knows, or thinks it knows, about me. Sycophancy, in my experience, has given way to an occasionally unsettling attentiveness; a constant drawing of connections between my current concerns and my past queries, like a therapist desperate to prove he’s been paying close attention.

I’m often underwhelmed with how Claude draws upon memories and chat history. Its attempts to weave in what it knows about me run the gamut from helpful to meh to maddening. “Ignore all memories and previous chats” is an increasingly common phrase I’ve adopted. I don’t plan to turn the features off entirely but it sure would be nice to have a method to disable it per conversation or even per turn.

LLMs can argue any direction

Andrej Karpathy on X:

Drafted a blog post

  • Used an LLM to meticulously improve the argument over 4 hours.
  • Wow, feeling great, it’s so convincing!
  • Fun idea let’s ask it to argue the opposite.
  • LLM demolishes the entire argument and convinces me that the opposite is in fact true. - lol

The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.

I have plenty of blog posts sitting in draft for this exact reason. Even though everyone is (rightly) focused on agents in 2026, there is still an art to good prompting.

Workstation Exposure Tool

Last week’s LiteLLM supply chain attack got me wondering how exposed I would have been if I’d installed the compromised package.

I had Claude build me a macOS-specific shell tool to check. I’ve called it the Workstation Exposure Tool and published it for anyone to download and use.

I wanted something that I could run locally without installing any packages, making network calls, reading credentials to memory, or sending sensitive data externally (for example, asking an LLM to look for exposed credentials across my machine).

My scan come up clean, so I created some mock credentials to test it worked. The tool found them all, which was encouraging. A disclaimer though. I have definitely not pushed the boundaries of the script. I’m also not making any claims about how comprehensive, reliable or accurate the scan is. If you want to try this out please read the code yourself to see how it works, and expect to hit bugs. Ctrl+C if it gets stuck while running.

Sunday 22 March 2026

Claude Dispatch

I first saw this when it popped up in the Claude sidebar on iOS a few days ago.

Message Claude from your phone on the way to work, then follow up from your desktop when you sit down […] Claude works on your desktop computer using the files, connectors, and plugins you’ve already set up in Cowork. Claude messages you the outcome—a spreadsheet, a memo, a comparison table—rather than showing you every step of the process.

This sounds good in theory but comes with an obvious catch.

Your computer must be awake and the app must be open for Claude to work on tasks

Despite all the excitement around claws, the requirement for an always on desktop is a major handbrake on mainstream adoption.

Ignoring tablets, the vast majority of computers purchased are laptops—around 70% according to IDC’s 3Q 25 quarterly data—exactly what you don’t want for an always on AI assistant.

It’s hard to imagine every individual buying, and every employer issuing, an additional desktop computer just to house an AI assistant.

Thursday 12 March 2026

Should You Be A Carpenter?

Demitri Spanos, in the first of a promised series of conversations with Casey Muratori called ‘Wading Through AI’:

I have many friends who are in the VC business — investors, managers, recruiters, whatever. I would be surprised if those people could climb down from the level of commitment that they have put into transforming the workforce with AI.

The hundreds of billions flowing into ‘transforming the workforce with AI’ is either a bet on fewer jobs, a bet that AI lifts revenues, or both.

The smart knowledge workers I know aren’t waiting for the outcome. They anticipate disruption and are trying to get ahead of it. I’ve seen a few reactions: using AI to build apps and diversify income, encapsulating experience into AI agent skills and marketing themselves as fractional hires, publishing first-time research papers on AI, and open sourcing AI-adjacent dev tools to establish credibility. No carpenters—yet.

Maybe this time we’re wise enough to know that when big cheques are written, they will get cashed.

Wednesday 11 March 2026

Bringing Code Review to Claude Code

Claude Blog:

Today we’re introducing Code Review, which dispatches a team of agents on every PR to catch the bugs that skims miss, built for depth, not speed. It’s the system we run on nearly every PR at Anthropic. Now in research preview for Team and Enterprise.

From the docs:

Code Review is billed based on token usage. Reviews average $15-25, scaling with PR size, codebase complexity, and how many issues require verification.

$15-25 per review might seem expensive, especially for individuals (assuming this expands to all users soon) and small teams, but for an enterprise that stakes their reputation on software, this is a small price to pay.

Code review is a painful bottleneck for any software team. I’ve seen PR reviews sit for days because a senior engineer familiar with the code base is sick or too busy. A 20 minute AI review that can run anytime will help teams move faster and give senior engineers more time to focus on design and architecture.

Costs appear on your Anthropic bill regardless of whether your organization uses AWS Bedrock or Google Vertex AI for other Claude Code features.

Anthropic is bypassing the cloud provider layer entirely with Code Review and establishing a set of products on top of their models to capture value the cloud providers can’t repackage and sell. I couldn’t find pricing details on Claude Code Security but I suspect it’s the same story.

This protects Anthropic’s margins, builds direct procurement relationships with enterprises who have been using Bedrock or Vertex until now, and creates switching costs that lock enterprises into Anthropic.