Tyler Folkman

CAIO at JobNimbus. I build AI systems that survive production and test what actually works. The AI Architect.

[Tyler Folkman @tfolkman]

@PeterLu98 @TechMDAI I just like the spark community more to be honest.

[Tyler Folkman @tfolkman]

Anthropic just filed its S-1. Headline: an $80.7B first-half net loss.

$72.4B of that is paper: accounting rules charge the company as its own valuation climbs, and the charge stops when the notes convert at the IPO.

The business itself lost $693M on $16.29B of revenue

[Tyler Folkman @tfolkman]

@TechMDAI Is that just due to the larger memory footprint?

[Tyler Folkman @tfolkman]

@austriker27 I built a chore app for my kids to give them rewards and manage their chores. Its been hit or miss, but it actually has helped on average.

[Tyler Folkman @tfolkman]

@ayayalar @TechMDAI Agreed . I just got one for 1295. You have to wait about a month for them to be delivered cause normally out of stock and back ordered.

[Tyler Folkman @tfolkman]

@plotarmordev @NVIDIAAI @ashxhart @u1tra_instinct Will do. Can you pick local voices for grok bot? I've not used it much honestly

[Tyler Folkman @tfolkman]

@ashxhart Gotcha - well if you ever need to run some stuff on 4, let me know. Im gonna try and contribute to 4x spark recipes now that I've got em connected.

[Tyler Folkman @tfolkman]

@tysn_dev DeepSeek v4.1

[Tyler Folkman @tfolkman]

@shotgunner101 Get 256-512gb ram and run DeepSeek v4.1 flash.

[Tyler Folkman @tfolkman]

@scottew I've not yet gone that far. I have a method to securely share logins but usually I just have it get things ready then I pay myself.

Might change with some of the agentic payments work happening from stripe and others.

[Tyler Folkman @tfolkman]

@majewskizby Quality rage bait. :)

[Tyler Folkman @tfolkman]

Looking for advice... Pi, Deepseek harness, or OMP?

[Tyler Folkman @tfolkman]

@ashxhart Nice work!

[Tyler Folkman @tfolkman]

I ran terminal bench 2.1 on my local quantized models and for me deepseek v4.1 1017 is the clear winner.

[Tyler Folkman @tfolkman]

This is the current lay of the land for open-weight vs closed at the frontier.

I am betting this changes in the next few weeks as we get some new releases with open weights.

Does anyone think open could actually be the top one day?

[Tyler Folkman @tfolkman]

What's not talked about on X that you wish were?

Can feel like everyone just talks about the same things: model release, local ai recipes, random comparison of models using three.js.

That is great and all, but what are we not discussing?