AI with Kyle Daily Update 154

Today in AI: Anthropic negotiates

What’s happening in the world of AI:

Highlights

Anthropic Back at the Table. Send Someone Else, Dario!

A week after being labelled a supply chain risk, Anthropic are back in negotiations with the Pentagon. Dario Amodei is holding discussions with Emil Michael, Undersecretary of Defense for Research and Engineering, to hash out a new contract.

Emil is…not a fan of Dario:

This is exactly what I said would happen. The TikTok playbook. Government throws a public tantrum, saves face, leaves a six-month window, and quietly sorts it out behind closed doors.

The sticking point hasn't changed: the Pentagon wants AI for mass domestic surveillance. Anthropic's red line is specifically that - domestic surveillance, not surveillance in general. They're fine with spying on non-Americans. Do remember that they are fine building lethal autonomous weapons as long as they can't harm American war fighters or civilians. This is from Amodei's own memo. Let's not elevate them as pacifists.

Here's the problem: Emil Michael - the bloke Dario is now negotiating with - called him "a liar with a God complex" last week. Publicly. So maybe… don't send Dario to the table?

Sam Altman navigated the same waters because he knows when to smile. Dario doesn't. Standing on principle is admirable. But if you can't cut a deal, Nvidia stops selling you GPUs, Amazon pulls your hosting, and Anthropic is finished. Game over.

Meanwhile, Amazon (Anthropic's largest investor and hosting provider) has said absolutely nothing. Conspicuously quiet. No defence. No statement. Honestly the silence is louder than anything.

Kyle's take: If a deal gets done, all that goodwill from the #QuitChatGPT crowd evaporates. People switched to Claude because Anthropic told the Pentagon to do one. When they realise Anthropic went right back and negotiated? That narrative collapses. Those ~1.5 million new Claude users might not stick around. Always remember these are billion-dollar companies run by billionaires. None of them are the “good guys”.

Google Workspace Gets a Universal Remote. This Is Massive.

This sounds boring. It is not. Well…it is. BUT it’s important!

Google released a command line interface for the entire Google Workspace: Drive, Gmail, Calendar, Sheets, Docs, Chat, Admin, the lot. One tool that connects AI agents directly into everything.

Previously, if you wanted AI to work with your Gmail, you needed browser automation, MCPs, APIs, workarounds, hacky solutions. Now your AI agents can talk directly to your entire Google ecosystem. No clicking around. No browser automation. Direct access.

Think of it like a universal remote. Five remotes on the table — TV, sound system, streaming, gaming. Replace them all with one that understands everything. That's what this is for Google Workspace.

I started using it this morning. Building a content management system where I drop an idea in, it automatically researches, writes a script, creates social descriptions, generates captions, and manages everything through Sheets and Drive. One ecosystem.

Previously I had this all patched together. And it worked. Most of the time. But this new CLI makes life a LOT easier.

Kyle's take: This is the boring plumbing that actually matters. Not flashy but super useful. If you use Google Workspace your AI agents just got direct access to your entire work life.

If you're not technical, you don't need to be. Tell Claude Code about it, paste the GitHub link, and ask what you can build with it. The AI will handle the installation.

NotebookLM Gets Cinematic Video. Still Free. Still Underrated.

Google keeps making NotebookLM better and nobody is paying attention. They've upgraded video overviews from static slideshows to full cinematic motion video, likely using their Veo model combined with Gemini?

The cinematic video is currently on Ultra plans only, but it'll come down to Pro and Free eventually.

Kyle's take: I spent two years telling everyone to use Claude. Nobody listened until the Pentagon drama. Now my next mission is NotebookLM. It's a mini RAG system that turns any information into any format.

Throw in sources, get out presentations, podcasts, study materials, video explainers. The moment they add NotebookLM to the new Workspace CLI, the content automation possibilities are truly astrounding. I'll be able to go from live stream transcript to finished multi-format content package without touching a thing….

Source: NotebookLM

Member Question from audience: "Is a Claude subscription worth it? I've only used ChatGPT."

Kyle's response: Claude is the better model. Full stop. But the pricing is a shock if you're coming from ChatGPT. Free plan is basically non-existent. £20 a month gets you into limits within a few Opus exchanges.

Realistically, you need the £100 plan for actual productivity. If you're running a business, that's the best hundred quid you'll spend on anything.

For personal use? It's a lot.

ChatGPT gives you all-day usage for £20.

Different companies, different strategies. OpenAI wants a billion users, most not paying. Anthropic wants a smaller group who pay well. Their revenues are roughly the same despite OpenAI having 900 million users versus Anthropic's 30-50 million.

Member Question from audience: "Where should I start with Claude Skills?"

Kyle's response: Start with your problem, not the tool. There's a directory with 200,000 skills - avoid rabbit holes like that!

Instead ask Claude for what you need. Say "these are the things I'm working on, are there skills that'd help?"

Go to anthropic.skilljar.com for free courses including one specifically on building skills. And remember, you can make your own, which is super cool. There's even a skill that helps you build skills. Very meta.

This question was discussed at [49:15] during the live session.

Streaming on YouTube (with full 4k screen share) and TikTok (follow and turn on notifications for Live Notification).

Audio Podcast on iTunes and Spotify.