@trykashu guy. AI & telecom Infra, energy, defense, manufacturing the rest of the time. Working to bring opportunities and people together for the greater good.
Argentina
Joined September 2022
- Tweets7.5K
- Following2.4K
- Followers1K
- Likes36.8K
Pinned Tweet
Major Considerations Dominating My Life With Little Detail:
1. AI Infrastructure: decentralized data centers (low mw), gas flare energy production, improved connectivity infra in Argentina, Morocco, and the US.
2. @trykashu specifically, improved security & control in payments
3. AI logistics and shipping infra & direct application. One company already has and is executing on dual use cases. Poised to be major.
4. Facilitating rapid acquisitions in manufacturing/defense & infra
5. BD bc what gets many companies to where they are won't get them to where they want to be.
Several lessons learned this year I want to go into.
Just sent the most unhinged “well it’s been some time since we last talked” to @JamesonCamp
Todays piece will be called “data is the new library of Alexandria” and I feel like people will not like it which is why I am posting it.
Yoda voice
Begun, the data broker wars have
agreed
real winners are:
-mercor/micro1/turing/surge/nyne/polyshares/grepped anyone with a lab contract
-anyone adaptable and good at making frontier tasks (task mercenaries)
-anyone with a proprietary data source (coding agents, personal agents, kled, microagi)
-vertical brokers can also sometimes make it (david/sftraces/sfdata)
Just Danny retweeted
Market update:
1. Labs are supply maxing - they are taking as much as they can and choosing what to close on.
This is the market in correction.
1 year ago from what I was told deals were going for $250k - same deal is around $100K now.
This is just the market on max supply right now.
This won’t last the markets next move to swing back to a more balance position. Prices won’t be at $250k but closer to $175k if that makes sense.
2. Labs are building their own GTM teams / paying referees 10%-15% of deal flow is starting to become “unsustainable” (they just want more of the margin). So they are building in house teams. Let me correct that - they are trying to build GTM, but their idea of GTM is Apollo and Lemlist / iykyk.
Meaning they will try and build GTM to maximize their margins but they have a ways to go. Not a threat to brokers or especially performance marketers. but nonetheless they are trying to replace you.
3. Insider told me that one of the big primes is moving to a fixed model for deals / no more negotiations on their offer/ pricing.
My thoughts is that this is part of the correction and once supply dies down - they will change that model again.
The reason supply will die down is once business owners and brokers understand that labs do not close nearly as many deals as people think - people will be less interested in getting involved. Reminds me of PE deal origination - look at 100 deals and maybe close on 2.
4. A lot more info on the battle that is happening between Frontier Labs and Data Labs… will talk about it in the future. Right now something is cooking….
Protect your data my friends…
Just Danny retweeted
More data brokering thoughts regarding where things are at, financing, and the like. View article
More data brokering thoughts regarding where things are at, financing, and the like. View article
Just Danny retweeted
deploying a swarm of catfish data broker hinge accounts to win the niche codebase acquisition game
Just Danny retweeted
me: have you tried selling your datasets to generate cash?
him: how would that work?
me: you have 66m datapoints. convert your raw unstructured data into rl tasks: user input -> agentic reasoning traces -> output -> evals
him: why would anyone want that?
me: labs will pay millions to train their models on this
him: you can't model ball knowledge
me: smh
Just Danny retweeted
If you believe that the market for data will 100x then it’s very likely data will become a tradable commodity.
As with any working market, liquidity is critical.
Liquidity for any asset class is unlocked when there is trust established around the contents without having to check firsthand. If you look at all the biggest markets, for commodities, energy, equities, and debt they cross the minimum threshold through grading and verification.
To move beyond one-off negotiations, there needs to be sufficient scale when it comes to buyer / seller transactions and for these to be a lot more open and public too.
I’ve sat down to refine or change things like ten times in the past few days but between actually being around business owners and working on deals I haven’t had a chance.
But something in some form is coming this week. Maybe case studies or something etc etc
Will be dropping my “The party is crowded and someone already puked in the punch bowl and why data brokering somehow got weird already” post tomorrow.
Also, I appreciate the folks with serious deals reaching out to me. I’m going to be a bit more curt in responses. I don’t mean to leave some in the lurch. I just don’t have a lot of time right now.
Just Danny retweeted
consistent feedback i've been hearing from big labs is that the existing data players are widely underwhelming. i don't think quality standards have dropped necessarily, it's more that we've moved past the era where contractors / experts alone can make data that is strong enough to challenge a model without being unfair. quality standards rise with model capabilities.
models are good enough now to synthetically, in-house generate tasks that are competitive (and infinitely cheaper) than SMEs. naturally there are exceptions in very unique fields, however, the broad majority (code, finance, law, etc), are all at a precipice where new players can come in and be boutique, quant-like firms that challenge the status quo.
these smaller players have: scalable systems that allow synth gen at scale; domain-expert engineers who understand both RL data & signal, and the field they're in well enough to design a system that can truly train a model; ways of creating heuristics for signal that can tell whether or not a task is useful in training; and real opinions where they propose the tasks, rather than responding to lab requests. they are more of research companies that come in with ideas, instead of being reactionary.
although there are many data shops that keep spinning up, in reality, the data space is only crowded, not competitive.
Will be dropping my “The party is crowded and someone already puked in the punch bowl and why data brokering somehow got weird already” post tomorrow.
Also, I appreciate the folks with serious deals reaching out to me. I’m going to be a bit more curt in responses. I don’t mean to leave some in the lurch. I just don’t have a lot of time right now.
The data brokering discussions I've had in the past few days cascaded into learning absolutely horrible truths about the industry but also connecting with some incredible people.
Imo you cant just blame AI for this, we all know why this is happening right now.
Call center jobs grew ~4% annually for 15 years
Now they’re shrinking 4% annually
Charts of the Week: a16z.news/p/this-week-in-cha…
Just Danny retweeted
I told someone earlier today who asked “can I just go message labs and find out what they’re looking for”.
I mean you can, but understand you’re one of many.
Why not find something you know they would want, then ask “hey I have this, would this interest you, if not, what are you looking for”
It’s not rocket science. Lead with value. Make others jobs easier, show initiative.
This also extends to asking about me or anyone helping facilitate selling your data. The more context you give and more you bring, the more likely someone will want to help you.
Alright, not WEIRD per se, but just got a deal lead from a friend of a friend.
Their family own over 60+ daycares for a very long time. They modernized when the sons took over from their parents about a decade ago.
There are certain industries I'm more apprehensive/sensitive to, then I realize better me than someone else.
And keep in mind this is just ONE buyer of data.
But you need to learn to be ahead and source what they really want.
Today micro1 is announcing a commitment to spend $1B over the next 12 months on the acquisition & licensing of enterprise data for AI training, with financial backing from Citi and Hercules.
micro1.ai/blog/micro1-commit…
