Jon Christensen

Kelsus CEO - We build sovereign AI document solutions for financial services companies.

[Jon Christensen @JonXtensen]

@psomkarbuilds Human emotion calibration vessel.

[Jon Christensen @JonXtensen]

@sunaiuse My 10 year old could have done this.

[Jon Christensen @JonXtensen]

@DavidSacks Play the rules on the field right? If you want attempts toward regulatory capture to be against the rules, perhaps you should convince your powerful friends to indroduce some regulation preventing it?

[Jon Christensen @JonXtensen]

Checking out a cave in Sardinia and pulled a quick wrist roll to keep my content game from going too dormant.

[Jon Christensen @JonXtensen]

Are investment bubbles a good or bad way for a nation to build its infrastructure? #aibubble #ai #industrialpolicy

[Jon Christensen @JonXtensen]

An open weight models ban? There’s one part that’s a good idea. #openweight #sovereignai #ai #softwareengineer

[Jon Christensen @JonXtensen]

The vocal, open-weight champions are saying to use open weight models for coding. That’s not what we’re saying. #openweight #ai #softwareengineer #aiautomation

[Jon Christensen @JonXtensen]

Arriba la celeste!! More watch movement prototyping. #watchmaking #watchtok #3dprinting #vamosargentina

[Jon Christensen @JonXtensen]

Doing some 3d printing of prototype movement parts. #watchtok #watchmaking #watches

[Jon Christensen @JonXtensen]

Ricky Skaggs picking tunes for a mostly over-65 audience in Beaver Creek yesterday. I only saw two other phones out the entire show. It was like stepping back in time. #bluegrass #livemusic

[Jon Christensen @JonXtensen]

For those tracking the closed source vs open weight model debate:

In an an open weight vs frontier model benchmark for financial document workflows. GLM 5.2 cost about about 7x less per million tokens than Opus 4.8, but the quality differences in document extraction and RAG were within the margin of error.

https://t.co/jMTlhPCdhN

[Jon Christensen @JonXtensen]

We've also run some benchmarks on these models.

In our case we did them for financial document workflows. GLM 5.2 about 7x cheaper/million tokens than Opus 4.8 but the quality differences in document extraction and RAG were within the margin of error.

Writeup below:

https://t.co/jMTlhPCdhN

[Jon Christensen @JonXtensen]

Nothing like finding out your friend’s business is your new competition 😂 #sovereignai #enterpriseai #openweight #ai #rag

[Jon Christensen @JonXtensen]

Only one open weight model caught the flaw in my benchmark procedure. #deepseek #ai #enterpriseai #sovereignai #softwareengineer

[Matthew Berman @MatthewBerman]

Chinese open source AI is coming for US enterprise

[Jon Christensen @JonXtensen]

How does Ornith 1.0 compare to GLM 5.2 on enterprise workloads? #ai #openweight #rag #enterpriseai #softwareengineer

[Jon Christensen @JonXtensen]

Open weight models work perfectly well when self-hosted in AWS and they’re cheaper per million tokens than Bedrock. #ai #enterpriseAI #openweight #GLM #softwareengineer

[Jon Christensen @JonXtensen]

Paul Kedrosky’s NYT Article “There’s One Clear Reason Americans Are Gooomy About AI” #ai #nyt #aiapocalypse #usagi

[Jon Christensen @JonXtensen]

Am I misunderstanding Dario here? #openweight #oss #ai #enterpriseAI #softwareengineer

[Jon Christensen @JonXtensen]

lets measure actions not words. If it doesn’t matter where it runs, then give us your weights. If it doesn’t matter whether it’s open, then publish the Fable 5 weights.

If your business models isn’t based on charging more for inference than it costs, then why are self-hosted inference costs so much cheaper for the same sized model?

We’ve benchmarked the open weight models on enterprise workloads and they’re very good and much cheaper.

[CG @cgtwts]

Anthropic CEO:

“open source is kind of a distraction you still can’t really see what’s happening inside the model, so it’s not truly “free.”

People should be paying a lot more attention to what Anthropic is doing.