Deep Insight Labs

1.7K posts

Deep Insight Labs banner
Deep Insight Labs

Deep Insight Labs

@DeepInsightLabs

Operating system for AI-native hedge funds @fdotinc

Katılım Mart 2025
383 Takip Edilen234 Takipçiler
Sabitlenmiş Tweet
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
Applications are now open for Podium’s founding AI-native PM cohort. We are selecting a limited number of serious builders, quants, founders, and emerging managers to test whether institutional-grade PM formation can be rebuilt for the AI era. Oversubscription expected.
Deep Insight Labs tweet media
English
2
0
5
679
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
@chamath You should consider substituting some of the workload with traditional ETL/ML/NLP pipelines. We estimate that around 30-70% of LLM calls can actually be dealt with by simpler models and techniques. seldon-ai.com/blog/how-much-…
English
0
0
0
41
Chamath Palihapitiya
Chamath Palihapitiya@chamath·
8090 works on production systems for large, often regulated, enterprises. Vibing isn’t tolerated because these are the systems that run western society - banking, power, healthcare, insurance etc. Over the last few quarters, the gains that we got from using frontier models inside of our Software Factory on these systems started to shrink but the costs kept doubling. This makes sense I guess, as in hindsight, we were initially asking the model to do mostly light work (generate basic PRs) and now we were asking it to do more complex work (mitigate dependencies across systems). Unless you grow context massively, be willing to run many A/B tests and iterate massively (ie use massively more tokens) complex tasks stay roughly unfinished by the model and requires the engineer to largely act alone. In other words, we find the last 5% (ie where a model is truly equivalent to a reasonable engineer) extremely difficult to achieve and extremely expensive to such a degree that the fully loaded cost of the model + the engineer will not pay for itself. So I asked our CTO to start thinking about other ways. We need our engineers to have access to the best tools BUT we also need to educate them to think even more for themselves - not less - in this last mile. At the same time, we need to find solutions that decrease our token costs by 90% - especially because these bleeding edge tokens are not nearly as cost effective as the tokens before it and are creating a big OpEx bill for us. I wonder how many engineers, in all orgs, are running amok right now by using the latest frontier models as a kind of slot machine. Increasingly turning their mind off, largely keeping productivity flat while their CEO and CFO deals with a massive token bill? My advice to you is that when you encounter this last 5% of very hard technical challenges in getting a complex system into production, be circumspect. The challenge of the last 5% is actually getting harder - especially as hundreds and thousands of code generation model runs run amok adding all kinds of random cruft into codebases that eventually need to be rationalized.
dnap@dnapway

Chamath reveals his company's AI token costs are doubling every 45 days but productivity is only up 5% "I sat down with my CTO today, I said how are we doing on token spend. And he said the most incredible thing, he said right now, our token costs are doubling every 45 days. I said well what is the downstream productivity? And he said maybe 5% max." "So my costs are doubling every 45 days, my upside is essentially flat. He said honestly, what we're finding out is that you need to use a lot more tokens to get to this next iteration of improvement because we've effectively already asymptoted." "We're going to take a step back and try to figure out what to do. I don't know how many other companies will actually go through this reckoning now, but the point is everybody in the next three or four years will for sure go through it." "I suspect that if you can get out now, you should get out now before all of that starts to seep into the water table. Because I think that's probably what allows you to get out at a huge price and raise a huge amount of money."

English
88
62
586
173K
Vansh
Vansh@vansh22b·
A @ycombinator partner (very active on X) asked me for $250k with a guarantee acceptance to their next batch😭 He said you'll raise few millions after batch anyways so 250k will be nothing. What happened to yc😭😭
English
106
10
640
314.2K
dale
dale@daleverett·
The music for this video was recorded live at @fdotinc by @daltonmeon :D How many databases does your AI agent need? > We think the answer is one. We built pgGraph (now downloaded by 1.5K+ developers with 500+ github stars) and combined it with Postgres and vector search inside @polygres . The managed beta is free with $50 in credits, link below. (ever seen an explainer video using manga? well now you have :D) We had dalton play the piano because there's just too much ai slop today.
English
12
12
52
3.1K
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
@Jason We can imagine that. Will token prices remain low indefinitely though? What would steady state pricing look like? I think those are the elephant in the room.
English
0
0
0
602
@jason
@jason@Jason·
When tokens go down 90% by the end of the year and then another 90% next year, everyone's opinions on artificial general intelligence and superintelligence are going to change radically I'm currently on an unlimited GLM 5.2 bittensor subnet and I can tell you your behavior changes radically when token prices plummet
English
260
172
3.4K
407K
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
GPT-5.6 is now live on Seldon. But not every repeat workflow should burn frontier tokens. Most AI apps start by sending everything to the best model. That is fine during MVP. But once the same extraction, classification, normalization, JSON formatting, or workflow-routing call runs thousands of times, you are no longer paying for reasoning. You are paying frontier prices for repeat application logic. Seldon is a drop-in LLM gateway that routes to GPT-5.6 when your workflow needs frontier reasoning, and compiles repeated calls into deterministic pipelines when it does not. One API. Two cost tiers. Zero refactoring. We are opening beta access with $50 in free credits. Comment “Seldon” below and we’ll send you the promo code. seldon-ai.com
Deep Insight Labs tweet media
English
0
0
0
96
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
Why should a pure router be a product and not simply a feature (which kind of exists already)?
Yuchen Jin@Yuchenj_UW

We desperately need a smart model router. 1. We’re seeing a model explosion: GPT-5.6, Grok 4.5, Muse Spark 1.1, GLM-5.2, and Fable 5 all launched within the past month. 2. Even for a single model family like GPT-5.6, there're 3 (Sol, Terra, Luna) and 5 reasoning-effort levels. That is far too many decisions for users to make manually. The best model should be selected automatically based on the task, latency, quality, and cost.

English
0
0
0
53
Yuchen Jin
Yuchen Jin@Yuchenj_UW·
We desperately need a smart model router. 1. We’re seeing a model explosion: GPT-5.6, Grok 4.5, Muse Spark 1.1, GLM-5.2, and Fable 5 all launched within the past month. 2. Even for a single model family like GPT-5.6, there're 3 (Sol, Terra, Luna) and 5 reasoning-effort levels. That is far too many decisions for users to make manually. The best model should be selected automatically based on the task, latency, quality, and cost.
English
233
100
1.6K
143.3K
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
@kathrynwu1 The "survive long enough" idea fails to account for opportunity costs. A 7-eleven has a very good chance of survival, but nobody would consider that a winner.
English
0
0
0
59
Kathryn Wu
Kathryn Wu@kathrynwu1·
When we graduated from YC two years ago, the message on the slide was simple: don’t die. Survive long enough, and your odds go up. That used to be enough. But now there are too many median startups that technically survive. The new bar is not “don’t die.” It is don’t accept becoming a median startup.
English
14
2
82
45.6K
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
Yeah, but it was only really just earlier this year that open source narrowed the gap with frontier labs so much (in practical and not benchmark terms). Got to give it some time. Enterprises move slower. The sharp increase in open source adoption on Open Router can be a leading indicator.
Ara Kharazian@arakharazian

I think Chinese and open source models are being overrated in terms of their revenue impact on OpenAI and Anthropic First, our latest Ramp AI Index shows Anthropic extend its gains in enterprise (42% of businesses use it). OpenAI is flat. DeepSeek at 0.3% of businesses.

English
0
0
0
29
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
GPT-5.6 is now available on Seldon. But the real question is: how many of your repeat LLM calls should still be frontier-model calls? A lot of production AI apps start by routing everything through the best model. That makes sense during MVP. But once the same extraction, classification, normalization, JSON conversion, or workflow-routing call runs thousands of times, frontier pricing becomes technical debt. Seldon is a drop-in LLM gateway that routes to GPT-5.6 when you need frontier reasoning, and compiles repeat workflows into cheaper deterministic pipelines when you don’t. For repeat applications, we aim to reduce LLM costs by 30–70% while improving latency, reliability, and auditability. We’re opening beta access with $50 in free API credits. Comment “Seldon” below and I’ll send you a promo code.
Deep Insight Labs tweet media
English
2
0
2
125
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
@1Umairshaikh Distribution. "Better" product is a subjective thing. Distribution and solving a real pain is the crux
English
0
0
1
7
Umair Shaikh
Umair Shaikh@1Umairshaikh·
why do some founders get rich with average products while smarter founders build better stuff and stay broke?
English
43
1
36
2.3K
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
@0xsachi No kidding. Still can't believe the OpenAI side event with K-pop girl group performance...😮
English
0
0
0
31
Miss Sentient
Miss Sentient@0xsachi·
ICML looking like couchella for ai researchers
English
3
1
11
758
Deep Insight Labs
Deep Insight Labs@DeepInsightLabs·
The economics of the model layer, barring a step shift in capabilities is commoditization. Think of how ISPs evolved from the dotcom era till now. People may have cared to pay a premium for a 256Kbps connection over a 56kbps one because it made a real difference (streaming multimedia content). Today, nobody cares if it's a 1Gbps or 10Gbps connection. The argument that gains in model performance compounds in economic output, just isn't true with the current capabilities and rate of advancements. Secondly, there are many layers (applications, processes, etc) in between the model and the real world that affects how the value translates beyond the control of these frontier labs.
zerohedge@zerohedge

Zuck: “The pricing from some of the other labs is very extreme and has very high margins. We think that there’s a real ability to be able to offer frontier or very high-level intelligence at a much more affordable cost.” Epic pricing war breaking out among agentic models.

English
0
0
0
45
Deep
Deep@DeepStarts·
"I only took 3 days to build SAAS with 0 coding knowledge using AI. Developers are obsolete." Their build:
English
419
3K
33.1K
2.8M
Michael Levin (YC P 26)
Michael Levin (YC P 26)@mike_drip·
biggest startup tip: tell people your product is still broken if they buy anyway you found a painkiller not a vitamin
English
10
0
43
2.9K
Celeste Ang
Celeste Ang@celesteanglm·
You can build from anywhere
Celeste Ang tweet media
English
15
0
42
1.6K
Lin Qiao
Lin Qiao@lqiao·
The most dangerous sentence for closed model providers is: "We switched models and nobody noticed." That's exactly what happened at @Gumloop. They replaced Opus 4.8 with GLM-5.2 across a company-wide agent and cut costs by ~5x. This is how platform shifts happen. The open alternative gets good enough. Then it becomes the default.
Max Brodeur-Urbas@MaxBrodeurUrbas

there are certain costs people have stopped questioning so when one of them drops 80%, the first reaction is disbelief open-weight models will give everyone that reaction at scale this year excited to be partnering with @FireworksAI_HQ on this

English
13
12
171
38.8K

Keşfet

@chamath @vansh22b @ycombinator @daleverett @fdotinc @daltonmeon @polygres @Jason