Todd Dailey

Here’s an Apple secret: Success comes from culture, not technology. Like many at Apple, I worked for many years under John Brandon, our longtime head of sales. He had his “Top Ten Rules for Success” which I carried around and lived by. It’s easy to spot someone from Brandon’s team by how they collaborate. I tried to be one of them. I mopped the floor (literally at times), I shared everything I knew without gatekeeping, I always assumed the customer knew their business. I think about AI the same way. Most AI advice right now is for builders. Almost nobody is helping the marketers, sales engineers, and execs who actually have to put AI in front of customers. That’s the gap I work in. I left Apple last month after 22 years. After some thinking and a lot of conversations, I’m now doing what I did inside Apple, but for other organizations: AI positioning, helping businesses understand AI, and creating Apple-level presentations and content.

@BrandonTahedl @alexocheema @doogie69 What company? There was the Virginia tech cluster, but I don’t recall large corporate clusters off the top of my head

I remember those days vividly, my uncle worked at a company that had the most of them installed at a single location, Apple came out and did an interview piece on the company, and posted it on the site. Apples Server Manager was epic, I still have my Mac OS X panther unlimited users key and cds. Amazing servers to work on, and the xserve raid was very, very sought after and competitive.

Apple used to make rack mount servers and storage, called Xserve and Xserve RAID. I was hired by Apple in 2003 to be an SE (now called Forward Deployed Engineer) 😀 on Xserve. A custom SKU for a server for AI isn’t something I see as possible, it’s too specialized. I could see: - more server-like features in the os. (In fact in the latest betas of Golden Gate there’s a feature to auto restart on power-on in the settings UI. A small affordance to server use.) - the old tower Mac Pro could be side mounted in a rack. I could see Apple making a rack mount for Mac Studio or Mac mini. (Or more likely reselling someone else’s product.) The OS gap is huge though, the default macOS security and privacy settings fight server use. A “make this Mac a server” checkbox in settings would be a huge win.

@technoHimbo Correct! I just don’t see that part coming to market

@twid P sure they have rack mount apple silicon now just not for not Apple

This idea of a new role solves a big issue in tech. Individual Contributors want to become managers not because they want to or are good at managing, but because they see it as a career path. At Apple you topped out at IC5 (some orgs have an IC6 (think “distinguished engineer”) but that’s rare. IC5 is equal to M2 which is “Manager” - so to move up the pay scale past IC5 you have to become a Sr. Manager (M3.) What this causes is manager creep, with way too many managers managing a few people. On the team I just left at Apple there were 13 people, and 5 of those were managers. 2.5 people per manager! This causes at least two problems: - the org becomes manager-led not expert-led, because managers of 1-2 people naturally gatekeep. - doing anything is immensely hard because there are too many approvals because there are too many managers. In the age of AI managers can easily handle more people, and with how fast AI is moving managerial friction is deadly to business.

@alexocheema @doogie69 Check it out! 😀 applemuseum.dk/apple_museum/u…

@twid I had no idea that was your start at Apple! Were you working with @doogie69 from the very start?

@alexocheema @doogie69 Yep! Doug and I were both supporting the original Xserve clusters. We sold a bioinformatics cluster in a box. The support email alias went to my inbox and the phone number rang my desk phone. 😀

One gateway, many profiles — profile-based message routing — A single multiplexed gateway sharing one bot token can now route specific guilds, channels, or threads to different profiles — each with fully isolated config, skills, memory, and secrets. Point your work Discord server at the work profile and your hobby server at personal, from one bot. A second multiplex hardening wave means one misconfigured profile can no longer take down the whole gateway.

@twid Profile routing? Could you explain?

I was literally just wondering if Hermes Agent had solid profile routing and was on my list to check, and they shipped it! ❤️

Protect your right to local AI!

Within an individual project that’s easy - you can just tell Claude or Anthropic or Hermes or openclaw to do it, you just need rules for what job runs where. Pareto is fun because your app defines a need and the model defines a capability, so as models change the model used by the code might change. I think this style makes sense long term. Openrouter can already arbitrage the cheapest model, so if deepseek is cheaper on Amazon versus google hosting it can pick that one. But yeah, I think baked in ai capability requests along with manifests on the model level on what they can do makes sense.

@twid I've observed the same model preference/behavior outside enterprise users! Any thoughts on baking in orchestration/request routing within config files of a project to have more control instead of relying on something like pareto router?

@0xSero I like the shape of this post.

At the beginning my load bearing instincts got wedged, but then I got to smoke-test my way to to THE smoking-gun, these belts and suspenders were absolutely right for the job. You'll hear the outcomes as they land.

@bookflywheel thank you!

@twid This was a very thoughtful and sharp analysis Todd! And you’re right the model routing layer is about to evolve into an entire new market opportunity. I suspect a lot of YC startups are going to jump in too 😅

@LyalinDotCom Same

I’m going to (maybe) live to 100 folks. 💪😅

@0xSero agentmail.to super good for this if you don't like your agent pretending to be you.

I'm realising the power of my agents talking to me friends agents. It costs very little for people to collaborate on complicated processes. Wow.

((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z) isn’t even that impressive of a polynomial all of those numbers and letters are in the training data

Reading list: OpenRouter's Pareto router: openrouter.ai/docs/guides/ro… @zachrattner 's Mac mini cluster: zachrattner.com/projects/m4-ma… Salesforce ambient listening: admin.salesforce.com/blog/2026/turn… @ollama 's MLX optimized Gemma 4: ollama.com/blog/faster-ge…

As I mentioned in a comment last week, the William Gibson sci-fi future of illegal software (AI models) trading in dark alleys may come to pass, it will just be much more boring and less stylish.

@saranormous So... a rubber goose? 😀

This is also how I justify purging my entire inbox periodically. They will email again 😆