Sabitlenmiş Tweet
Kaba
1.8K posts
Kaba
@kabalabsinc
if attention is all you need, train yours
Sync Katılım Ağustos 2023
195 Takip Edilen244 Takipçiler
much discussion of automating processes in the enterprise that are hugely manual as the killer app for agents.
One Big Problem: these processes are opaque and live among all sorts of half baked and out of date data
those tasked with writing out the supposed process usually fail
English
@nbaschez “80% of the world’s data is private…”
public internet been slurped up, they need the pathway that leads to the eg reddit/twitter post on niche topic by the chemist
Troy Kirwin@tkexpress11
yesterday we had @kevinweil (VP of Science Open AI; fmr head of product at IG and twitter) come hang with the @speedrun founders one thing he said jumped out: >> we asked: "what startup opportunities do you find most interesting?" >> Kevin: 80% of the world's data is private. OpenAI won't have access to it. Getting access to that data (or creating data) and building a business on top gives you a real moat.
English
The number of new wildly fast growing companies in the emerging "training data for AI labs" industry continues to surprise me
Has anyone written anything about how this market works and why so many new companies keep coming out? Do the labs intentionally wanna fragment their suppliers?
Spencer Mateega@spencermateega
For most of history, expertise was scarce, constrained by time and reach: one person, one career, one lifetime. Now, for the first time, we can encode, evaluate, and scale it. We believe the wisdom that once took a lifetime to build shouldn’t take a lifetime to find. Today, we’re excited to announce that @AfterQuery has raised a $30M Series A at a $300M valuation and that we’ve since surpassed $100M in annual revenue run rate, to build the data layer of professional AI.
English
@matt_slotnick not trying to one up, bc you are correct, but also fun to have an unintelligible LI account and watch striver brains explode trying to parse it
English
Have
Too
Many
Tabs
Open?
G
O
V
E
R
T
I
C
A
L
!
Google@Google
Too many @GoogleChrome tabs open? Try vertical tabs, rolling out now. Just right-click any Chrome window and select “Show Tabs Vertically” to move your tabs to the side of the browser window, making it easier to read page titles and manage tab groups.
English
Kaba retweetledi
I’m quite sure people don’t Get It yet, but this literally is the game.
>Do you have trusted access to edge environments you control that induce the origination of scarce context valuable to commerce?
If you do - yay you captured something scarce. If not 🥲🥲 you’re a middleman
illiquid@lefttailguy
This may not seem like a big deal. But if getting to ASI is increasingly a function of 1) inducing the origination of scarce context, and 2) access to the permissioned environments where that context is likely to be originated, a lack of trust becomes existential for OpenAI. Progress is increasingly downstream of trust.
English
You can just go to Church Hayek.
Peter is implicitly making the case for @lefttailguy’s Middle Game, tho it sure looks like this middle is the end.
Scarcity economics is not that complicated: Own the context flows that uniquely unlock commerce and you’ve found thiels nirvana
Oguz Erkan@oguzerkan
Peter Thiel: “AI chips will get commoditized.” Yes, the value in AI has largely accrued to the semiconductor level so far amd concentrated largely in $NVDA. But this is bound to change. In chips, you can stay as a monopoly only if you have an exclusive partnership with the monopoly in a downstream market. This is how $INTC dominated CPUs for decades, by partnering up with $MSFT. We can’t see such a partnership in AI, industry dynamics don’t allow it: - Downstream market is fragmented. - Price/performance drives decisions. As a result, the more AI workloads shift to inference, the more fragmented the AI accelerator market will become as inference-specific chips or custom chips of hyperscalers will be used more. Yes, it looks like most of the value has accrued to $NVDA right now but it’s almost certainly not sustainable. This is why I don’t own it despite seemingly attractive valuation. Its base is already too high and it’ll certainly struggle to maintain it let alone grow.
English
@hypersoren best part is no one asks how this hyperlegible real time intelligent queryable state happens… saas apps + llm? OS? Throw it all in a datalakehousebase? vendor canvas/workspace?
all wrong answers
English
Taking a company as an assemblage of its context flows and its access to markets running on commodified intelligence you can quickly see why the coasian singularity is ridiculous on its face
Minimizing transaction costs now means reducing context revelation + transacting less!
Brian Halligan@bhalligan
NEW EPISODE: @jack & @roelofbotha unpack @blocks 40% staff cut and rebuilding the entire company as a mini-AGI. This isn’t “use AI to make people more productive.” It’s making the company itself the intelligence. If you’re a founder or operator wondering what work looks like in the next 5 years… this is the episode. The evolution looks like: • Manager mode = Pyramid 🔺 (command & control) • Founder mode = Flat ➖(founders decide fast) • Dorsey mode = Circle 🔵 w/ AI at the center, humans at the edge, and decisions flow from customer inputs → AI → humans steering it I’ve tried killing org charts before. Brutally hard. But we never had these tools. This is rewriting the CEO playbook for the AI era. Buckle up. 00:00 Existential Dread & Hope 02:56 AI Replaces Hierarchy 07:22 Block’s New Three Roles 26:47 Flattening the Company, Fast 35:23 Getting the Board to Buy-In, Fast 36:50 Building a Great Board 41:29 Founder CEO Lessons 48:18 Second Acts & Conviction 56:22 Timeless CEO Traits
English
@bhalligan thesis antithesis synthesis
very relevant as ai will combine aspects of eg on prem and cloud into something new
rethink it all… arch, biz model, all of it
English
I got a chance to interview Jack Dorsey for the Long Strange Trip pod about his new article (in comments) on how to build a modern ai native company in 2026.
I took some of his ideas and mixed them with some observations about the hyper scaling ai native CEOs I work with every day. This is a first draft. Thoughts? If this is landing, I can flesh it out more.
English
copilots happened before cockpits. dumb.
Jarvis only exists after the sensors, networking, data, hud, training, and deploying test/shadow model to that same environment for real time rl
Kaba@kabalabsinc
@ashebytes the human is the loop the sensor is the actuator the harness is the sandbox the attention is the weight the model is the product everything must be created bottoms up and node (person) out
English
More converging ⬇️
Kaba@kabalabsinc
@akoratana @JayaGup10 this is part of a broader convergence. when computer security _is_ alignment: 1) enterprise security, observability, tooling will come to consumer 2) the enterprise will need real time query/policy enforcement of associate endpoints 1 converges b2b + b2c, 2 cloud + local
English
@km next token prediction on someone else’s computer is saas/subscription… wrong biz model for ai
Kaba@kabalabsinc
@tmrohan @michaelonx_ new tech => new delivery mechanism => new biz model(s) this incongruence is wildly under discussed next token prediction on someone else’s computer is just saas
English
Still hard to understand, and true for the last 50 years. Just have to find the right ones.
Terrence Rohan@tmrohan
We’re pricing tomorrow’s companies against yesterday’s business models. Seed is under-priced.
English
@JayaGup10 @amritwt unironically, we will have our own search/page rank/recommender algos
mostly used for discovery & search of unique, expert weights
English
Consumer built empires by turning every click, hover & abandon into rocket fuel.
Agents changed that.
Decision traces just became the new compounding loop.
Old feature-moat SaaS is cooked.
Trillion-dollar rewrite loading…
Jaya Gupta@JayaGup10
English
@ashebytes the human is the loop the sensor is the actuator the harness is the sandbox the attention is the weight the model is the product
everything must be created bottoms up and node (person) out
English
a theme many of you have posted on:
the agent attention tax
you’re the human in the loop with your agentic systems (coding, second brain brainstorms, briefings) but you’re in too many loops and it’s exhausting.
the product edge this year will be evidence of continual learning - where you trust that the agent system is actually removing you from number of loops it requires over time while still feeling tightly coupled & aligned to you
English
@josiezayner running agents (read:spam bots) to create accounts and run ads through.
bet on it
English
Everyone wants this to be true so bad count down until we find out its not actually real
The Rundown AI@TheRundownAI
Sam Altman predicted in 2024 that a one-person billion-dollar company "would have been unimaginable without A.I., and now it will happen." He just emailed the NYT saying he won a bet with tech CEO friends over when it would arrive, and that he "would like to meet the guy." The guy: Matthew Gallagher, 41. Spent $20K and two months building a GLP-1 weight-loss telehealth company out of his living room in LA. The stack: ChatGPT, Claude, and Grok writing code. Midjourney for images. Runway for video ads. ElevenLabs handling customer calls. Custom AI agents stitching it all together. $401M revenue in year one. On track for $1.8B this year.
English
@akoratana @JayaGup10 this is part of a broader convergence. when computer security _is_ alignment:
1) enterprise security, observability, tooling will come to consumer
2) the enterprise will need real time query/policy enforcement of associate endpoints
1 converges b2b + b2c, 2 cloud + local
English
Never thought about it this way before...B2C companies had the self-reinforcing loop figured out a long time ago, and it's coming to B2B just now
Jaya Gupta@JayaGup10
English
autonomous agents presented as an opportunity for software but it’s absolutely not
you think companies are going to let middleware turn them into paid rl env to rent back intelligence at replacement costs?
get real
logan bartlett@loganbartlett
English