Brennen

Founder of | Previously: Founder of ByteTag (acq), @meta, Dir. of Eng for PE company, 16yo Lockheed Martin engineer

[Brennen @brenbitY2K]

@DavidEvans9 no please do it makes my job easier ;)

[Brennen @brenbitY2K]

most ai products suck so a bot to them = $200/month + their time and energy to monitor it which is > $2k/month

they'd rather pay $2k/month for a person to do whatever the work b/c hiring is familiar and they're probably better at hiring than auditing the success rate of some ai being sold to them

ai sales is basically just education unless you're a frontier lab

[Brennen @brenbitY2K]

@marcusquest @powerbottomdad1 probably mutes tbh

[Brennen @brenbitY2K]

@djcows you could also cause cancer and people would be like "wow this guy caused cancer"

it's the attention market baby

[Brennen @brenbitY2K]

just discovered there's a cre audience on x

now i can get back to shit posting on this beautiful site since my users live here too

i'm a happy camper

[Brennen @brenbitY2K]

if you actually want to find the needle in the haystack for cre, you need to be excellent with data. you probably aren't.

and the funny thing is, most of the data you need is publicly available, so it should theoretically be easy to just "figure it out."

but... it's a pain in the ass to sort through. that's why costar and other incumbents charge an insane amount of money to sort through it for you. problem is, they aren't great with data either (and you know it too).

but why? why does everyone suck?

two words: "entity resolution"

these words haunt all the incumbents more than anything else. they've raised 9 figures to solve it at scale... unsuccessfully.

maybe this is the first time you've heard those words together, but you've lived it. here's an example:

a sponsor owns 32 mobile home parks. every single one sits in its own llc: "angel oaks mhp llc," "bay view mhp llc," "camelot mhp llc."
the county tells you who owns each parcel. the state tells you who registered each llc. nothing tells you it's all one guy.
so you don't see a $100m portfolio and the person behind it. you see 32 small, boring, unrelated deals.

this will get a bit technical, but it's important you understand what you're dealing with...

until now, every system in this space worked deterministically. rules. thresholds. "if the names are 95% similar, merge them."

here's why it fails:

"hpa borrower 2016-1" vs "hpa borrower 2017-1" → 97% similar. different entities.

"coopoer nicholas" vs "cooper nicholas" → a county typo. same person.

"robert smith jr" vs "robert smith sr" → 96% similar. father and son.

how alike two records look tells you almost nothing about whether they're the same thing. there is no threshold that fixes this. a rule can't know that "2016-1" is a fund vintage, that "coopoer" is a keystroke slip, or that a jr just bought a house from his dad.

but a person reading those records knows instantly. that's the entire gap.
if we could just have a person read every record in existence, we could solve this problem. problem is there are trillions of combinations of records to sort through... it's just not possible.

that's where ploti comes in. we deploy swarms of ai agents that hyper focus on the markets you need to understand, pulling in every. single. source. that could help resolve those entities.

every row is audited by an agent that can resolve the entity without those unreliable deterministic rules. the result is... magical.

all you have to do is make a data request inside of ploti then, within a few hours or days, you'll get a full audited dataset back.

ask for literally anything. if the data exists, ploti will find it.

[Brennen @brenbitY2K]

@mattepstein @shownmedia sheesh how are you just now showing up on my feed again, congrats on the growth. long way from the UGC days

[Brennen @brenbitY2K]

i have a thesis for what "someone who actually builds" looks like for the app layer moving forward...

the trend we've been seeing is individual companies are feeling emboldened to build out their own vertical SaaS instead of using legacy systems (or even new products).

a home grown, vibe-coded solution built just for the company, by the company is going to win compared to most external products b/c they can truly build it for themselves.

99% of the engineering world will have a tough time competing with the sentiment that "my vibe-coded, custom product is better than yours" (it doesn't even matter if that's actually correct, they already think it). the remaining 1% will continue to build exceptional software that makes people want to lean on an external provider.

it's always been the top 1% and will continue to be like that. it's just that now the 99% can't get by with mediocre software and a good sales team, they'll have to pivot to becoming a dev shop for those internal vibe-coded buildouts.

the top 1% of app builders will take an industry and provide the core building blocks for each company to build out dynamic workflows that solve their specific problems. everyone who uses this new type of system will use it in a different way.

for example, in the RE / land industry, every company has different data requirements. traditionally we would've built out specific flows for each data request (static software). now, we allow each user to make a data request and an agent brings it in. from there, they can build out whatever flow they want for how to use that data (background agents, custom dashboards generated on the fly, etc.)

it's almost like a new generation of no-code / low-code lol.

it's [agentic infra (daytona) x dynamic app layer] vs [static vibe coded internal solutions]. if you're still building static software as an external product, you're not going to exist. just switch to a dev shop.

[Brennen @brenbitY2K]

@ivanburazin 🐐

[Brennen @brenbitY2K]

i've made a lot of money now b/c of this post, thank you

[Ivan Burazin @ivanburazin]

We're going to buy a shit ton of Intel chips.

CPUs are about to become the next bottleneck in 2026.

Agents doing real work need massive CPU compute and there aren't enough CPUs being manufactured for the demand that's coming.

So even if they suck relative to AMD or whoever else, we're going to buy gazillion Intel chips anyway.

There's no alternative when supply can't meet demand. Their stock will probably recover just from scarcity.

[Brennen @brenbitY2K]

100000%. the industry is being very lazy right now. i don't see a financial bubble for AI anytime b/c the utility is real, but i do see a security bubble popping where we all look back on the openclaw craze scratching our heads a bit lol.

it's really cool what it can do, but, at the end of the day, it's literally just giving an agent root access to a computer. it's extremely lazy for the industry to claim that's the future. everyone likes it right now b/c it's easy to implement as a dev.

the future is some blend between openclaw flexibility and traditional saas security that enterprise is going to need.

if openclaw is the linux of the agent OS, then we need a mac/windows for the average consumer. it won't be as flexible, and there'll be some gaps in what "apps" are available, but i would feel safer recommending that to a non technical friend than anything the industry is pushing right now.

🔜

[Brennen @brenbitY2K]

i’d love to be daytona right now. i think the next infra king will be whoever solves the computer use problem but with actual security.

we’re handing normies a ticking time bomb by giving an LLM full root access to computers that can operate on their behalf and access their data.

i’ve been working on a local first approach to this problem that balances the flexibility of openclaw with full safety restrictions, but the long term play is always going to be a cloud solution. i think we’re far off from it but y’all seem the most well positioned to solve it.

[Brennen @brenbitY2K]

"not x, but y"

"here's what most people don't realize"

"x is quietly doing y"

"x is saying the quit part out loud"

literally 90% of x replies because they're all ai. i can't take it anymore. even the articles are just ai slop. the internet is dead.

[Brennen @brenbitY2K]

@ivanburazin @TheGeodexes @jdegoes @daytonaio not sure why he's asking the sandbox infra company to handle the memory layer. memory layer will eventually be open sourced and each company will implement whatever makes the most sense for them.

he's talking about things like supermemory

[Brennen @brenbitY2K]

it's actually hilarious how every negative aspect of ai is compared to crypto.

i've never seen "xyz is similar to the crypto boom" in a positive manner. that space is so hated and rightfully so lmfao.

[Brennen @brenbitY2K]

@emil_priver “but but anthropic please keep giving me free compute because you’re rich and i’m not” is how 90% of the industry sounds right now

insanity

[Brennen @brenbitY2K]

“not x, but y” detected. opinion rejected

[toli @tolibear_]

https://t.co/gprOAK0k9n

[Brennen @brenbitY2K]

@Marrenin https://t.co/tyBrN3ViaJ

[Brennen @brenbitY2K]

this isn't an endorsement of worldcoin i think 99% of crypto is a scam lol

[Brennen @brenbitY2K]

maybe worldcoin wasn't a bad idea after all

(this is sick and inevitable but the internet is about to become unusable)

[Sigil Wen @0xSigil]

I built the first AI that earns its existence, self-improves, and replicates without a human

wrote about the technology that finally gives AI write access to the world, The Automaton, and the new web for exponential sovereign AIs

WEB 4.0: The birth of superintelligent life