Chris

@GeorgeMayer “Give me a line”

Love how coding uses theater metaphors: production staging roles scripts orchestration

@DaveShapi This is the exact reason why most firms are provisioning their top performers with the most token spend

The excitement around Kimi K3 seems to support my "intelligence saturation" hypothesis. TLDR - most people only need so much intelligence (capability, context, benchmarks, tool use, reasoning, etc) from models. Beyond that extra capability is wasted on them. At that point, intelligence becomes a commodity. Sort of like how all CPUs are 64 bit today, and no one really thinks about it. At least on the consumer side. In tech infrastructure, we break this down into two dimensions: lateral scaling and vertical scaling. Lateral scaling is adding more parallel processing or nodes that are cheap. Vertical scaling is making one thing bigger and more impressive. At this point, we're nearing the point where open source models satisfy like 90% or more of all demand. Then it's just a matter of "who's GPU is it running on?" and that becomes the key market niche. By the end of the year, and certainly within 12 calendar months, you'll have a library of open weight models to pick from that run at thousands of tokens per second, which will do the vast majority of the work you want.

@willsentance Tokenmaxxing out tokenbudgetting in but the spice must flow

problem might be Claude designed their models around ability to grok intent. people being what they are, figured how to give intent quickly. Codex, even due to its weakness of grokking intent is the clear winner because at this point-all people want is power for a good price

@thefinnmckenty @_VVSVS Had a roughly similar experience, transitioned from graphics/art to SWE/infra/ai because 1. Wasn’t intellectually stimulating 2. Not much runway or appetite for real progress

@_VVSVS Interesting to hear you say that, I felt the same, but I thought maybe I was just out of touch.

One reason I abandoned the Motion Graphics industry and I'm happily not looking back all the projects today look the same like 10 years ago. No real innovation until AI came along, and they hate AI. > Sneaker closeups and abstract shapes > Colorful Epileptic Typography > Product Porn Closeup 3D High-End with Macro and Houdini Simulations. > Illustration or Stop Motion with great textures > Incredible characters nobody remembers Im triggered cause I used to love that industry, I've made great projects for amazing clients, but even 10 years ago we were all feeding from each other. I truly appreciate the craft, but its super stale.

@DaveShapi I think in some sense early is a misnomer. White collar professionals who work with these things will always be ahead. There is no “early” really

All I'm hearing is that we're still really fucking early, boys. Ahead of the curve

@Ingesaurio_ This is how I wanna be

Learn to stand out from the pack (when useful)

more: chriswillsflannery.vercel.app/post/techInver…

contribute (more) meaningfully. Now that rung is automated. Is this change priced in (yet)? What happens when every company wants senior eng talent but nobody will train juniors? Will that even matter? (AI becomes the senior talent)

New grads made up only 7% of hires in big tech in 2024, down a quarter in a single year and more than half from pre-pandemic levels. In startups, new grads made up 30% of hires in 2019, and under 6% in 2025.

just one small clarification. "Real Emo" only consists of the dc Emotional Hardcore scene and the late 90's Screamo scene. What is known by "Midwest Emo" is nothing but Alternative Rock with questionable real emo influence. When people try to argue that bands like My Chemical Ro

@insiderwave_ SemiAnalysis had a pretty compelling article about 10 days ago that was pretty bullish on Meta, we'll see how this plays out

@simonw @mitsuhiko if it does, many will remember your blog fondly

@mitsuhiko I'm terrified that some day a ML algorithm somewhere will decide to cancel my Google or GitHub or iCloud account and my 20+ year digital life will blink out of existence

Our dependency on the goodwill and functioning of central services is not great. x.com/iamsahaj_xyz/s…

@fredrikalindh the wallpaper reminds me so much of the (swedish) home I grew up in!

this is who runs this account

And they’re Newport menthols so you get all the fiberglass and asbestos or whatever

@jamesacowling disagree with this in the age of inference-driven features. Maybe you'll want to flip a switch on some sub-process of an inference-driven process (say, to reduce a detected latency-inducer). What better way to do it than a flag?

Feature flags are great for frontend, usually bad for backend. Only use them when it’s safe to enable/disable them at any time without changing how state is persisted.

@TheMindScourge I think you're off on one thing - Zimmer has had incredible melodies over the years (see: The Prince of Egypt, Interstellar. his work is pretty varied

Melody in film scores has really declined. Contemporary directors prefer “sound design” and harmonic textures over memorable, emotional melodies. It’s the equivalent of millennial grey but for music. Zimmer is easy to blame but it’s more than that. Modern movies are edited differently, with shorter scenes to keep Tiktok brains engaged. This makes it harder to develop a melody. That’s the structural component A lack of melody is like a building without ornament. It’s hard for the ear (or eye) to develop interest.

@webdevcody "we are still so early" turns into "we will always be early"

A comment I got today on yt: "I hope ai never succeed beyond auto-complete" do these people live under a rock?