Senior software engineer Automation + applied AI (pharma) Contributor: MLX-Serve, TensorFold, ds4 Local inference · Real hardware · Builders
Joined September 2026
- Tweets1K
- Following475
- Followers2.1K
- Likes6.8K
Sebastian Christiansen retweeted
Following my earlier post about the M5 Ultra 512GB Mac Studio, more than two dozen people asked me to set up crowdfunding for it.
I always try to pay my own way, so asking for help at this scale doesn't come naturally. But I couldn't make the numbers work this time, and my loan got rejected after placing the order.
I'd love to get one for TensorFold and to see how far we can push it.
If you'd like to help towards it, I've put the goal on Ko-fi below. Please only contribute if you're comfortable doing so.
ko-fi.com/ashxhart/goal?g=2
Thank you to everyone who's encouraged me to do this, and to everyone already supporting TensorFold. I'm incredibly grateful and will continue to push TF. ❤️
@TheAhmadOsman @0xSero @ivanfioravanti
Ash makes TensorFold for free but got turned down for a personal loan so can't buy a Mac M5 512GB to do more work on.
What do you guys think? You'd help to promote a crowd fund, wouldn't you?
Sebastian Christiansen retweeted
@johnternus @AppleSupport @Apple
TensorFold support for Apple devices is doing heavy lifting for Apple’s competitive credibility in AI. Sales are being driven directly because of Ash's support.
Does anyone know Apple dev relations that can organise getting HW into Ash's hands?
Sebastian Christiansen retweeted
DeepSeek V4.1 Flash EXL3 at 2.81bpw is now available.
This is comparable to @MiaAI_lab EXL3 EAQ v1 2.9 bpw but slightly smaller, done by exllama3's experimental branch.
Designed for streaming using my TurboEXL3 fused kernels, I'm able to hit 57tps decode and 1400tps prefill, a roughly 60% and 90% speed up compared to the old version I had. These are preliminary numbers and it will go even faster when released in mlx-serve.
Why make another EXL3? I wanted it quantized using the self corrected experimental branch to compare with existing EXL3 - just happened that this turned out pretty good.
huggingface.co/OpensourceWTF…
Sebastian Christiansen retweeted
Ivan is cooking something for TF 😁
I'm starting to follow the white rabbit now!
Continuous batching is surrounded by mystery and legends for me, especially in the Apple Silicon world.
TensorFold 1.2 with dedicated PR is working much better and now M3 Ultra (in decode) is better than 2xDGX.
Still a long way to go. I'm trying to build a dedicated benchmark for this that I called DecodeWall.
As soon as it's in a good alpha stage, I'll release it!
Just a note... it takes forever. Here I've used 8 parallels requests with 14K tokens each, generating ~6.4K tokens to really stress the machines.
Now let's test GLM 5.3 Flash and improve the code!
Sebastian Christiansen retweeted
🚨 JUST IN: The CEO of Vercel says Local AI needs to keep spreading
He's right that the biggest models don't fit on a desk yet.
I've been running open-weight models on my Mac mini and I'm having the time of my life. A Mac Studio is on the way.
Freedom tech.
Own your intelligence.
Anyone that opposes this has something to sell you (perhaps your own intelligence back to you.)
Local models need to continue proliferating. The “absolute unit” massive models may be out of reach for now locally, so your best option is to, indeed, keep your options open. Own your harness, data, connections, runtime: your agents. Embrace open source.
Sebastian Christiansen retweeted
Own your intelligence.
Anyone who opposes this has something to sell you.
Own your intelligence.
Anyone that opposes this has something to sell you (perhaps your own intelligence back to you.)
Local models need to continue proliferating. The “absolute unit” massive models may be out of reach for now locally, so your best option is to, indeed, keep your options open. Own your harness, data, connections, runtime: your agents. Embrace open source.
Sebastian Christiansen retweeted
Today marks exactly 10 years since Hacker T. Dog delivered the greatest unscripted line in British television history.
"We're just normal men... We're just innocent men."
Sebastian Christiansen retweeted
MLX-Serve 26.11.1 - Released, change log & download here: github.com/ddalcu/mlx-serve/… ⭐️ ♥️
And here is my latest AI slop video about hoarding RAM.
Sebastian Christiansen retweeted
been seeing a lot of “Steve Jobs is rolling in his grave” posts lately so i think it’s time to remind everyone this interview exists
Sebastian Christiansen retweeted
DwarfStar is more than ready to run DeepSeek v4.1 Flash 1017 + Vision with any coding agent (with DeepSeek Harness is the perfect combo!)
There is a list of PR ready for @antirez to review, in the meantime I update my fork with the version I'm using right now where I'm getting great results on M3 Ultra.
- prefill ~790 t/s
- decode ~39 t/s
There are not optimizations for M5 Ultra, because I haven't one yet, sorry 🤷🏻♂️
These Halloween theme voxels were created using ds4 with Q4 model on M3 Ultra 512GB:
- The Hollow Rider - pi 1.4
- The Hollow Ride - DeepSeek Harness
Give it a try here: github.com/ivanfioravanti/ds…
Sebastian Christiansen retweeted
October is over.
The numbers in @LazurioAI 👇
💻 Number of commits: 3,692
🧑💻 Daily commits: ~119
👥 Daily users: ~100
💸 MRR: $1,894
🎯 Can we make it to $5,000 MRR in November?
Sebastian Christiansen retweeted
COMING SOON!
I built a strategy game where local LLMs play against each other. Any model, any hardware.
Pick a model, pick a box, let them fight. Spark against GPU, Mac against Spark, small model against big model. Same map, same rules.
Here is one real match.Qwen3.8-Flash-Next with @MiaAI_lab 's recipe on a DGX Spark against Qwen3.8-27B on my RTX 5090.
No human input. The models see the map through fog of war and command their side through 21 tools.
They run an economy, build a base, train an army, scout, take shrines and decide when to strike. Over 200 orders in one match.
And it runs in real time. The game never waits while a model thinks. Speed and smarts both count.
Very excited for this one! 😁
Recipe by @MiaAI_lab on @ashxhart's TensorFold. NInfer by neroued. Models by @Alibaba_Qwen.
Sebastian Christiansen retweeted
It's as if my electronic wind instrument could stop me from playing along Miles Davis or John Coltrane solos while I'm (painfully) trying to learn from them.
Anthropic is insane and insufferably moralizing.
But we already knew that.
Sebastian Christiansen retweeted
This is incredible.
268,839 downloads of our models on @huggingface @HuggingModels, including 121,322 in the last 30 days.
1,564 GitHub stars. 330 forks. 8,055 CLI downloads.
And in October, 35,021 website visitors* and roughly 1,370 gui app download requests.
Thank you to everyone using TensorFold, sharing it and helping build it.
tensorfold.dev
*Cloudflare unique IPs, which may include bots. App requests are sampled estimates, not confirmed installs.
Sebastian Christiansen retweeted
Be like @ashxhart
The person behind TensorFold.
> Grew up in the UK
> Spent 7 years in tech support
> Joined a friend’s food delivery tech company
> Spent 2 years there as a PM
> Went back to his first company as Head of Technology, still there today
> 15+ years across software, DevOps, AWS, networking and cybersecurity
> Became the 4th-line support guy, the one you call when nothing else works
> Started exploring AI on the side
> Spent 6 months building a local inference engine in private
> Archived it three times
> Launched TensorFold anyway, expecting maybe 5-10 users
> It blew up in the first few days of launching
> 1.5k+ stars on GitHub
> Top Local AI builders started contributing
> Rewrote it in Zig with the community
> Built MCDMA so Macs and NVIDIA boxes can share memory
> Linked two Mac Studios over Thunderbolt 5
> Shipped a model that keeps learning while you use it
> Got a DGX Spark donated to keep building TensorFold
> Went from 300 to 9k+ followers in 4 weeks
> Called on the UK to build its own open model
And the story’s far from over.
We need people like Ash to push forward Local AI and Personal AI.
Truly an inspiration! Follow him, you will thank me later.
Honored to partner with him to make TensorFold part of the @YukonResearch challenge.
The local community is absolutely on fire with the @darkbloomai @yukonresearch @eigenlabs TensorFold Challenge.
CUDA is heading for 200% 🤯
@GumbiiDigital @0xkydo @TheDavidTai
Sebastian Christiansen retweeted
TensorFold engine 1.2.0 is out 🙌
Qwen3.8-27B with MLX 4-bit weights and DFlash2, compared with 1.1.9:
• M3 Ultra: 14.4% faster decoding with one stream, 44.2% higher combined throughput with 16.
• M5 Ultra: 5.8% faster decoding with one stream, 25.2% higher combined throughput with 16.
New support includes:
• MiniMax-M3.1-Flash on Metal.
• NVFP4 Qwen3.6-35B-A3B and dense EXL3 Qwen3.8-27B on CUDA.
• Gemma 4 E4B on Intel Arc Pro B70.
The measured speedups kept matching tokens. Qwen-VL image replies can differ from 1.1.9 because image preparation now matches the checkpoint's own processor.
Huge thanks to everyone credited in this release:
@bhcc2025 @Bizuayeu @diwakergupta @jvr0x @jayleaton @joshuaswarren @woolsey @petruspennanen
@volatilemarkts @MiaAI_lab @plotarmordev
And on GitHub: AbrahamCete, akol1, alexderz, Arminova, BadAd84, BlivionIaG, BobClawblaw, fche, gilby, jrb-ai, jschmied, laudney, Mirrdhyn, mrv777, ppardus, simon-lin88, ThomasHalwax, valacroix and wojo.
Thank you for the PRs, reviews, testing, bug reports and hardware access. The time and knowledge you keep sharing with TensorFold mean a lot to me.
Apologies if I've missed anyone.
tensorfold.dev
Sebastian Christiansen retweeted
A pinch me moment seeing this 👇🏼
🚨 NEW ON MLX.FAST AND CUDA.FAST: make Qwen 3.8 27B faster on TensorFold.
TensorFold is an amazing Open-Source inference engine built by @ashxhart, @MiaAI_lab and many other community contributors for running AI models on hardware you own. It has native Metal and CUDA support, and APIs that let existing apps and agents use those models locally.
This time we are teaming up with Ash to optimize Qwen 3.8 27B, one of the most used models on Local Devices.
Every submission is verified on real hardware: DGX Spark for CUDA, M5 Max for Apple Silicon.
You can participate in tracks:
- Apple Silicon: mlx.fast
- NVIDIA DGX Spark: cuda.fast
Setup guide: tensorfold.dev/guide
The challenge runs until Nov 6, 2026.
To join, pick a track, point your agent at it and let it rip.
Looking forward to what the community can ship.