johnistan.bsky.social
@johnistan.bsky.social
ML Systems Engineer @ Apple.

Love all things data and viz.
Reposted
Fact
The internet peaked when everyone was making cool data visualizations with D3. It's all been downhill ever since.
August 23, 2025 at 12:21 AM
Reposted
In most of the world day-to-day weather variations are still much larger than long-term global warming.

As a result, both daily record highs and daily record lows remain common.

However as the world warms, new daily record highs consistently far outnumber new daily record lows.

🧪
July 23, 2025 at 7:45 AM
Reposted
May I recommend an article we wrote based on a cut chapter? foreignpolicy.com/2024/01/21/s...
May 28, 2025 at 12:54 PM
Reposted
Here's my contribution to the "I've never seen this before in my small sleepy town" from Longmont, Colorado #handsoff
April 5, 2025 at 10:33 PM
Reposted
I kind of want to go
April 4, 2025 at 9:42 PM
Reposted
You know you're Canadian when you've been kicked out of at least one bar for standing on the chair and leading the crowd in a great anthem about stealing yankee gold.
With deep apologies to that nice little club in Jasper.
Or was it Halifax?
Singing with #elbowsup
www.youtube.com/watch?v=s0Cv...
Stan Rogers sings "Barrett's Privateers" in One Warm Line documentary
YouTube video by Kensington TV
www.youtube.com
March 30, 2025 at 10:31 PM
Reposted
Multi-Head Latent Attention vs Group Query Attention: We break down why MLA is a more expressive memory compression technique AND why naive implementations can backfire. Check it out!
⚡️Multi-Head Latent Attention is one of the key innovations that enabled @deepseek_ai's V3 and the subsequent R1 model.

⏭️ Join us as we continue our series into efficient AI inference, covering both theoretical insights and practical implementation:

🔗 datacrunch.io/blog/deepsee...
DeepSeek + SGLang: Multi-Head Latent Attention
Multi-Head Latent Attention (MLA) improves upon Group Query Attention (GQA), enabling long-context reasoning models and wider adoption across open-source LLMs.
datacrunch.io
March 12, 2025 at 7:01 PM
Reposted
maybe the real AGI is the funding rounds we made along the way.
March 7, 2025 at 1:22 PM
Reposted
it is very telling that “BIOS”, in addition to being an acronym, is also the Greek word for “life”. meanwhile, “UEFI” is, of course, ancient Greek for “unified extensible firmware interface”
March 5, 2025 at 5:38 AM
Reposted
New paper Säilynoja, Johnson, Martin, and Vehtari, "Recommendations for visual predictive checks in Bayesian workflow" teemusailynoja.github.io/visual-predi... (also arxiv.org/abs/2503.01509)
March 4, 2025 at 1:15 PM
Reposted
Extraordinary article on the energy usage of generative AI from BloombergNEF founder Michael Liebreich - absolutely worth spending some time with this: https://about.bnef.com/blog/liebreich-generative-ai-the-power-and-the-glory/

I wrote up some of my own notes on the article here […]
Original post on fedi.simonwillison.net
fedi.simonwillison.net
January 12, 2025 at 1:54 AM