Seth Pratt

4.5K posts

Seth Pratt banner
Seth Pratt

Seth Pratt

@sethprattsf

what’s sota this week?

San Francisco, CA Katılım Kasım 2011
2.1K Takip Edilen1.3K Takipçiler
Seth Pratt retweetledi
Dr. Heidi Klessig
Dr. Heidi Klessig@heidiklessigmd·
Ms. Suzanne Chin, a lawyer and mother of two, suffered a heart attack in 2009 while she was living and working in Hong Kong. Despite intensive efforts at resuscitation, she remained comatose and was declared brain dead. Doctors advised her husband to remove her from life support. “Then, three days after she was admitted, she woke up from her coma. She recovered within a week and left the hospital. According to publicly available information, Ms Chin is now living in Singapore, still working as a lawyer, still a wife and mum. She is well, and she is alive.” There is a simple medical explanation for recoveries from “brain death” such as Ms. Chin experienced: global ischemic penumbra, or GIP, which is like a power outage of the brain. During periods of low blood flow, the brain (like any other organ) shuts down its functions to save energy. When brain blood flow decreases below 50% of normal, the brain becomes quiet and unresponsive to testing, exactly mimicking “brain death.” But brain tissue destruction doesn’t occur until brain blood flow drops below 20% of normal for several hours. People in the 20-50% GIP range of brain blood flow will not respond during brain death testing, but with continued efforts to improve cerebral blood flow, their condition is potentially reversible. GIP is like a power outage in your home: nothing works, but the wiring isn’t destroyed. Get the current flowing again and the lights will come back on. In the same way, with continued efforts to improve cerebral blood flow, the lights will come back on for some of these patients as well. Unfortunately, most don’t get this chance because brain death is a self-fulfilling prophecy: these people very quickly either have their support withdrawn or become organ donors.
Dr. Heidi Klessig tweet media
English
41
590
2.1K
104.7K
Ava
Ava@noampomsky·
friend is in the stage of claude psychosis where he asks claude to send him newspapers about what claude is doing for him
Ava tweet media
English
158
232
4.7K
171.5K
Seth Pratt
Seth Pratt@sethprattsf·
@vec0zy Theirs is slow as hell, temu version openclaw. Much cheaper though.
English
0
0
1
425
cozy
cozy@vec0zy·
ouhhh… husbandt… you spen 2k on maca mini, say you have special agents, stay up very late, no build nothing, so late for work too many time. now anthropic build special agent, cost no mronies… can just use shell phrone. now we are homeress and im so cold…
cozy tweet media
Thariq@trq212

We just released Claude Code channels, which allows you to control your Claude Code session through select MCPs, starting with Telegram and Discord. Use this to message Claude Code directly from your phone.

English
31
60
1.9K
254.9K
am.will
am.will@LLMJunky·
Finally proud to announce that I've joined the GPU Minor Leagues. 2 x RTX 6000 Pro. I have six months to pay off the second GPU lol. You are all TERRIBLE influences.
am.will tweet media
English
119
14
884
46.1K
Seth Pratt
Seth Pratt@sethprattsf·
@AlexFinn you look at how much those studios are worth on ebay now? I saw one sell for $26k last night. crazy.
English
0
0
0
4
Alex Finn
Alex Finn@AlexFinn·
OpenClaw and Hermes agent on the right, Crimson Desert on the left Multiple agents autonomously building businesses while I play the sickest video game ever made This is the future Your AI employees go out and create value while you enjoy the finer things in life I love 2026
Alex Finn tweet media
English
155
21
660
38.6K
Alex Cheema
Alex Cheema@alexocheema·
@danveloper Close. We will see someone running a 1T model on $16,000 of heterogeneous hardware (Apple Silicon + NVIDIA) at 100tok/sec this year.
English
7
1
77
6.2K
Dan Woods
Dan Woods@danveloper·
We’re gonna see someone running a 1T model at 100tok/s on a $2500 laptop by like a month from now. Long Apple, this is the best AI inference platform.
English
26
21
546
38.4K
Seth Pratt
Seth Pratt@sethprattsf·
@TheAhmadOsman yeah DGX spark has a very narrow use case, not useful for practical work flows as far as I can tell. like an inferior mac studio. not enough tokens/s for most tasks.
English
0
0
0
132
Ahmad
Ahmad@TheAhmadOsman·
DGX Spark uses unified memory > 273 GB/s RTX PRO 6000 delivers > 1.8 TB/s (1792 GB/s) If someone told you they’re comparable, they’re wrong And this is exactly why llama.cpp isn’t the right tool here Try vLLM or SGLang on a GPU and you’ll see very different results
Max Weinbach@mweinbach

@TheAhmadOsman I have on DGX Spark and then was having insane tool calling issues and was told by Nvidia to use llama cpp

English
40
13
313
32.9K
Seth Pratt
Seth Pratt@sethprattsf·
@maxyourtest Hear me out here…. Just go to a gym and sit in the sauna every day? Same result, no parasites
English
1
0
9
1.9K
Take Testosterone
Take Testosterone@maxyourtest·
New libidomaxxing method just dropped
Take Testosterone tweet mediaTake Testosterone tweet media
English
190
342
9.4K
1.6M
Seth Pratt
Seth Pratt@sethprattsf·
@AlexFinn you have just saved me so much money in API costs. holy cow. 70% of my per token cost was dead cron data being loaded into the context every message.
English
0
0
1
61
Alex Finn
Alex Finn@AlexFinn·
IF YOU'RE ON OPENCLAW DO THIS NOW: I just sped up my OpenClaw by 95% with a single prompt Over the past week my claw has been unbelievably slow. Turns out the output of EVERY cron job gets loaded into context Months of cron outputs sent with every message Do this prompt now: "Check how many session files are in ~/.openclaw/agents/main/sessions/ and how big sessions.json is. If there are thousands of old cron session files bloating it, delete all the old .jsonl files except the main session, then rebuild sessions.json to only reference sessions that still exist on disk." This will delete all the session data around your cron outputs. If you do a ton of cron jobs, this is a tremendous amount of bloat that does not need to be loaded into context and is MAJORLY slowing down your Openclaw If you for some reason want to keep some of this cron session data in memory, then don't have your openclaw delete ALL of them. But for me, I have all the outputs automatically save to a Convex database anyway, so there was no reason to keep it all in context. Instantly sped up my OpenClaw from unusable to lightning quick
English
205
110
1.8K
274.1K
Seth Pratt
Seth Pratt@sethprattsf·
@ideabrowser wouldn’t be hard to do this with a couple of open source local models on Mac. need a bit of unified memory to work with though to keep everything up. Not sure how you could pull this off without a decent gpu on windows though.
English
0
0
0
17
Idea Browser
Idea Browser@ideabrowser·
Startup idea that customers are begging to buy right now. Open Granola -Same meeting capture (no bot joining calls) -macOS Native desktop app -Notes stored as markdown files. -API (not MCP) -Local-first, or privacy. - Agent-readable by default. -Plays nicely with ai-first tools. Granola just locked down their local db, went MCP only. Someone is going to build a more open version and print money. If you build this... 1. I'll buy it and be your first customer. 2. I'll help you launch and get users. DM me the link.
Guido Appenzeller@appenz

Sorry to see Granola @meetgranola going closed. They encrypted their local db, no local and no cloud API. In a world where notes are managed by agents, the app now has zero value. Any recommendations for good alternatives? What are you switching to?

English
67
7
243
48K
Dan Woods
Dan Woods@danveloper·
I handed Claude Code @karpathy's autoresearch repo and Apple's "LLM in a Flash" paper, told it to get Qwen3.5-397B running on my M3 Max 48GB... it did!
Dan Woods tweet media
English
84
166
2.4K
206.6K
Rake
Rake@GoForRake·
@SterlingCooley I have my US 1000 right beside my red light panel, so I'll give this a try right now. I'm sitting in place for 20 minutes anyway, so may as well do both!
English
2
0
1
162
Sterling Cooley
Sterling Cooley@SterlingCooley·
This research is actually more important that I initially realized* We may be able to target abdominal regions (no more spleen location - which is difficult) if this research holds up and expands further, I don't see why it wouldn't scale up, this could be massive for US-1000 and US-2000 Pro owners, you guys are so darn lucky ! Recent research confirms that transcutaneous focused ultrasound applied to the epigastrium suppresses systemic inflammation by modulating subdiaphragmatic vagus nerve fibers. This non-invasive approach utilizes acoustic pressure to stimulate mechanosensitive vagal afferents. In controlled rodent models, a 10-minute application at 1 MHz resulted in a 180% increase in vagal firing. The physiological result is a significant downregulation of pro-inflammatory cytokines. Data shows a 48% reduction in TNF-α and a 42% drop in IL-1β within two hours of stimulation. The mechanism targets the cholinergic anti-inflammatory pathway. By increasing efferent tone, ultrasound signals the spleen to inhibit cytokine release via α7nAChR receptors on macrophages. Human feasibility trials demonstrate a 30% cytokine reduction. This bioelectronic intervention offers a precise, drug-free alternative for managing rheumatoid arthritis, IBD, and sepsis. pubmed.ncbi.nlm.nih.gov/41671184
Sterling Cooley tweet media
English
7
36
176
7.3K
Seth Pratt
Seth Pratt@sethprattsf·
@TukiFromKL Nvidia has been paying dividends and doing buybacks for the last several years
English
0
0
0
652
Tuki
Tuki@TukiFromKL·
🚨 Do you understand what Nvidia just told you? The company building the backbone of AI - the one spending $10B on new fabs - just said "we don't know what to do with half our money." That's not generosity. When a tech monopoly in the middle of an arms race starts handing cash back... they're telling you the growth story is over before anyone admits it.
unusual_whales@unusual_whales

BREAKING: Nvidia, $NVDA, says it is planning to use 50% of free cash for investor returns

English
25
26
444
138.1K
Seth Pratt
Seth Pratt@sethprattsf·
@IndieGameJoe I worked at blockbuster for a month, why anyone would want to do this as a game is crazy to me.
English
0
0
0
128
Indie Game Joe
Indie Game Joe@IndieGameJoe·
Two indie devs made a game where you run your own video store in the early 90s. It’s currently the #5 top-selling game on Steam. - Rent out VHS tapes & manage customers - Charge Late & Broken Fees - Upgrade & customise your store It’s called Retro Rewind - Video Store Simulator
English
1.3K
4.7K
53.7K
13.6M
Matthew Berman
Matthew Berman@MatthewBerman·
What’s the best open source model for coding right now? Any size 😆
Matthew Berman@MatthewBerman

.@nvidia hand delivered a pre-production unit of the @Dell Pro Max with GB300 to my house. 100lbs beast with 750GB+ of unified memory to power the best open-source models in the world. What should I test first?

English
73
9
185
31.8K
Seth Pratt retweetledi
Peter Steinberger 🦞
Been so much fun cooking OpenShell and NemoClaw with the @NVIDIAAI folks! 🙏🦞 Huge step towards secure agents you can trust. What’s your OpenClaw strategy?
English
246
215
4.3K
221.8K
Seth Pratt
Seth Pratt@sethprattsf·
@Midnight_Captl or buy a Mac Studio 512gb and a dual 5090 for similar power at a quarter of the price
English
0
0
0
339
Midnight Capital LLC
Midnight Capital LLC@Midnight_Captl·
I just found out 30 minutes ago there was a product announced that nobody knows about cuz it wasn’t in the keynote It’s called the DGX Workstation, and it’s a single GB300 for your home It’s $100,000 and Nvidia’s calling it a one time purchase of a digital employee 🤯 $NVDA
English
19
5
170
18.6K