JW (jay,dub) (@builderjaydub) header
JW (jay,dub) (@builderjaydub) profile picture

JW (jay,dub)

@builderjaydub

founder @heurist_ai building agents for finance

 New York
Joined May 2013
  • Tweets4.6K
  • Following3.5K
  • Followers7.1K
400Media
Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)Photo by JW (jay,dub) (@builderjaydub)
what's the last time you had croissant with pasta and wine?
Image from JW (jay,dub) (@builderjaydub)
Muse (@Muse) profile pictureMuse
@MuseNov 6
Replying to @Muse
Tonight’s space special: every restaurant you've saved on Instagram and never went to, with prices, ratings, and more 🍽️ These are just a few ideas to get started, but there is plenty of space(s) for more. What will you build?
1
1
346
i'm sure this pizza is tasty
Image from JW (jay,dub) (@builderjaydub)
70
Datacenters can be positive sum
Shoeshiner Capital (@madridraptor) profile pictureShoeshiner Capital
@madridraptorNov 6
1/ Texas is now arguing over whether it wants AI data centers at all. So I looked at a county that already has a giant one: Dickens County, pop. ~1,700, home to $GLXY Helios campus. What does the local public record say? A sourced thread.
2
463
I don't know how much lower we go in AI There are a lot of reasons to be skeptical of overbuild and cyclicality But let's keep in mind... - agentic adoption really started in Q3 and is an order of magnitude more token demand than an individual person (always on, stackable) - frontier cyber AI only recently became public an is now a permanent tax on enterprises and nations (and is escalatory in nature no ceiling) - the rest of the world, including China, is behind the US in their buildout and in penetration - speaking of penetration, Anthropic's S-1 rounded to 0.1% of enterprise spend on AI - that's a low amount for intelligence as a service! - robotics & autonomy are still sub-scale but real
3
2
2
34
4,182
Buybacks in q4 going to lead a strong bounce take my word
Beat (@beat_markets) profile pictureBeat
@beat_marketsNov 5
Lam Research $LRCX announced a $25 billion stock buyback hours after falling 16% on its worst day in a year.
243
Anthropic has a new agent scaffold besides Claude Code, reported by @METR_Evals "People can spam 10 ideas at the scaffold for autonomous experiments" - said a researcher My guess is that the internal agent scaffold is the main driver for RSI
Image from JW (jay,dub) (@builderjaydub)
2
1
1
495
This possibly explains why Claude Code team members left recently. Coding is no longer the frontier
1
80
worth a read in the middle of $QQQ $SOXX crash
METR (@METR_Evals) profile pictureMETR
@METR_EvalsNov 5
Since August, a team within METR has worked with internal Anthropic information to assess whether AI itself is accelerating AI development. We are now publishing a report summarizing the results of this risk assessment.
357
JW (jay,dub) retweeted
Anna Wong (@AnnaEconomist) profile picture
Anna Wong
@AnnaEconomistNov 4
If you think the AI trade is collapsing after today, and data centers delivery next year will be pretty much stalled, you would think that 10 year should be substantially lower than 4.9%, cuz we’d be talking about sub-1% GDP growth, possibly even talking about the R-word (which i am specifically banned from saying). But if not…. Then what world are we in?
Image from Anna Wong (@AnnaEconomist)
98
102
21
847
101,067
JW (jay,dub) retweeted
josh :) (@jfonsecarivera) profile picture
josh :)
@jfonsecariveraNov 3
some people downplay this, some overhype it but i'm excited for one reason: this work is verifiable and we've seen what happens to verifiable domains (hint: a few years ago, models struggled with 8th grade math)
Rikiya Takehi (@rikiyatakehi) profile pictureRikiya Takehi@rikiyatakehiNov 2
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Image from Rikiya Takehi (@rikiyatakehi)
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
12
13
6
134
5,294
Let GPT 6 Astra hand-write weights of a 1.6m param model to predict bitcoin price I just asked for the weights. It beats the baseline without training
Image from JW (jay,dub) (@builderjaydub)
2
373
Replying to @kennyistyping
are you serious right now? all our stuff is open source and fast. that's why the devs use them and love them, and people cant rmb how it was to dev on ethereum before we build these tools. you need the devtools to build anything. the devtools gotta be fast else your devs churn. 101. basics! promoting the usage of a slow, painfully dumb setup e.g a miniature quantized qwen in your qubes over tor using kohaku is not making the world a better place or promoting a helpful ideology, it's making you sound like a fucking hipster, and worse you're baiting other people with potential in dumb hipster philosophies.
11
9
2
153
8,448
don't you like the halloween costume?
Image from JW (jay,dub) (@builderjaydub)
Georgios Konstantopoulos (@gakonst) profile pictureGeorgios Konstantopoulos
@gakonstNov 3
it's kind of painful to see the devcon keynote after almost 10yrs building ethereum's OSS stack, is this what is left? where is the accountability? how are you going to inspire the next wave of strong devs? not the B players and the rent seekers? in the age of AGI? come on
356
$GPRO R/R looking good here. 1 $GPRO = 1.14 cash post merger + 0.1 share of the new photonics company 15% of float is short. A squeeze may happen any time if retails buy it up
Image from JW (jay,dub) (@builderjaydub)
1
1
685
We love the camera @ShaneStoffer @markiplier
67
JW (jay,dub) retweeted
bread.mega (@bread_) profile picture
bread.mega
@bread_Nov 2
Tired of these DPRK hacks. So I built the ultimate defense. DPRKaptcha[.]com Simple questions with existential outcomes (for them). Live now to protect the internet from North Korea.
Andy (@andyyy) profile pictureAndy
@andyyyNov 2
THIS IS INSANE. In just 4 days following the biggest exploit ever, Mercado Bitcoin is hacked for $40M and an EU exchange is hacked for $106M today. Is anything safe anymore??? 🤯
Image from Andy (@andyyy)
Image from Andy (@andyyy)
165
426
126
4,212
353,780
JW (jay,dub) retweeted
Chandan Singh (@csinva) profile picture
Chandan Singh
@csinvaNov 2
For anyone questioning whether this post is overhyped, I think they are actually *underselling* the approach. We did similar experiments and found that Claude models can write entire, functional neural net weights from scratch! csinva.io/weight-space-progr…
Rikiya Takehi (@rikiyatakehi) profile pictureRikiya Takehi@rikiyatakehiNov 2
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Image from Rikiya Takehi (@rikiyatakehi)
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
99
251
65
3,193
298,674
Most trading apps are built for attention-grabbing and make you over-trade. I vibe-coded my own terminal with the right amount of information. Custom layouts, watchlists, minimal set of indicators (MA, Fib, sigma levels) that I care about. Bye Tradingview!
Image from JW (jay,dub) (@builderjaydub)
Image from JW (jay,dub) (@builderjaydub)
3
1
785
Combining with Claude + @heurist_ai workbench, I can now analyze every names of my book, let AI create a trading plan and show it up in the terminal. nitter.gq/heurist_ai/status/2113…
Heurist (@heurist_ai) profile pictureHeurist
@heurist_aiOct 23
Heurist Finance workbench is now open source. Bring your own agent and start working like a pro investor.
1
107
frontier model capabilities blowing my mind
Rikiya Takehi (@rikiyatakehi) profile pictureRikiya Takehi@rikiyatakehiNov 2
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Image from Rikiya Takehi (@rikiyatakehi)
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
6
497
Load more