They're adding valueβ’ And they're very proud of it. @BragsVentures
SF
- Tweets 21,262
- Following 4,826
- Followers 305,462
- Likes 193,930
βIn a world of AI slop, the only real moat is taste and thinking for yourself.β
- Spencer, 26, Stanford CS dropout, YC W26, building a personal assistant AI agent, lives in a Hayes Valley hacker house, into peptides, sauna socials, and posting about "new media"
6
2
66
3,036
humblebrag /ΛhΚmbΙlbrΓ¦Ι‘/ noun
A statement that expresses pride in an achievement, possession, or circumstance while appearing to complain or be modest.
Example:
19
8
706
30,730
Turns out we can predict the future
excited for what comes next
Our prediction for this Prime Day comes from Mimi, our newest prediction model at Social Intelligence Labs.
Mimi simulates 60 million people using census data and enriched persona data from other public sources and social media. We have been building Mimi over the past few months and are excited to test it publicly for the first time.
Overall, here are our thoughts:
A smaller event than Prime Day. About $109 per household, roughly a quarter below June's $143, with orders around $47 (June: $48). About half of households place two or more orders (June: 63%).
Awareness high, enthusiasm modest. 91% knew about the event, but only about one in three (32%) shopped mainly because of it, and 56% are very or extremely satisfied with the deals.
Careful shoppers. About one in three (32%) are more cautious because of the economy, and 54% compare prices across
retailers.
Early holiday shopping, light AI use. About one in four (23%) buy holiday gifts; about one in five (21%) use AI shopping tools
We will report back when Numerator's data comes out in a couple of weeks.
1
2
12
5,991
VCs Congratulating Themselves πππ retweeted
Genuinely canβt tell if this is a typo because itβs a huge win for Europe either way
Nvidia will announce $28 of AI investments in Europe this week $NVDA
65
323
13,784
364,874
Ground breaking investment for the European economy.
The sheer volume of money being put behind AI investments in the EU putting America to shame
Nvidia will announce $28 of AI investments in Europe this week $NVDA
4
76
10,876
VCs Congratulating Themselves πππ retweeted
VCs will watch a college football locker room speech and think it's exactly what the founder of a B2B expense management startup needs to hear.
4
2
57
8,399
Five people, two weeks, $300k
Apparently you can build a frontier AI model without raising $500 million and hiring half of your employees from Stanford and Ivy League universities
Inconvenient development for the rest of SF
We are releasing Ultramarine 1 Flash, a frontier-class model built for knowledge work. It is open-source and achieves similar performance to Fable 5, while being 60x cheaper.
It was post-trained by a team of 5 for ~$300k all-in. Available on API now, with weights releasing in 2 weeks.
Paid partnership (ad)
16
2
439
66,264
VCs Congratulating Themselves πππ retweeted
Venturemaxxing Night Recap
140+ VCs & LPs, one warehouse turned into a museum: legendary brags duct-taped into gold frames, the fleece vest, the holy trinity of every VC bookshelf, a few dead startup toys (RIP Juicero), and the iPhone with ventureβs most brutal text βweβre oversubscribedβπ
Thanks to everyone who filled the gallery with aura and to our co-host @VCBrags, and a very special thanks to our partners @sydecario, @harmonic_ai, @Google, @Stifel, and @GoldenEggMediaβ€οΈ
Next stop is NYC in November - donβt miss it!
1
24
3,102
The device in question
We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and weβre grateful to the agents and prosecutors who moved quickly to make an arrest. The FBIβs investigation is ongoing and we will continue to support their work.
At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated.
We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.
3
2
80
9,737
VCs Congratulating Themselves πππ retweeted
enough of these cliff hanger posts trying to engagement bait people into guessing who you invested in
17
6
366
17,379
Does this mean I can clone a famous angel investors voice to hear about they invested early in Uber?
$52M seed round. 15M+ users. But this is what we're most proud of yet.
Introducing Drama 3: the most emotive and controllable TTS model in the world
And yes, it's open-weight.
You can now control how even a single word should sound, change its tone, and shift your voice mid-sentence.
Uncle Roger was impressed too.
Try it here:
s.fish.audio/rqh0l8
Comment βDRAMAβ within the next 48 hours for 1 month free.
If you're an enterprise your first month is on us:
s.fish.audio/qz9r0t
Paid partnership (ad)
3
49
12,315
when a VC starts posting startup advice but they've never built anything in their life
Billy M
6
8
279
20,188
Imagine if Jian Yang could handle up to 384K tokens of context back in the day. Not Hot Dog might still be around today
Introducing SubQ Snap, the first long-context decision model for real enterprise work.
Snap is the only decision model or LLM where compute per token stays constant as context grows.
β’ 384K tokens of multimodal context, 12x larger than Jev.
β’ 61.7 on Decision Index v0.2.1 vs Jevβs 57.9, outperforming on 13 of 16 enterprise benchmarks.
β’Processes 100K tokens and returns a decision in under one second.
β’$0.05/MM input tokens. Output is free.
Available today to anyone.
The first 10,000 sign-ups get 100M free input tokens.
Weβll likely hit capacity fast and will add users as we add compute. Compute offers welcome π.
Latency will be variable as we tune the platform to this new model and category of users.
Model speeds are 80ms for minimal context, 250ms for 32k, 660ms for 100k, and 2.5 seconds for 384k.
Visit: subq.ai/
Paid partnership (ad)
7
28
9,434