Pinned Tweet
OpenAI launched their Forward Deployed Engineering agency today. We've shipped 30+ custom AI agents and tools in the past six months.
Here are 12 agents and AI tools we've built for clients:
1. AIEO visibility platform that monitors brand presence across LLMs, generates high intent prompts to track, and delivers execution plans.
2. Sales call roleplay trainers where reps pitch voice AI prospects trained to talk like their real buyers. Personalized, post-call coaching slashed onboarding time and boosted win rate.
3. AI GTM agents that build prospect lists, research every account, and write personalized messaging. Creative campaigns that used to take days now takes minutes.
4. Lead scraping and enrichment agent that processes tens of thousands of leads, scores by problem symptoms, enriches contact info, and syncs to CRM. Sellers open their laptop to qualified leads every morning.
5. AIEO/SEO content agent that pulls customer insights, keyword research, and proprietary data to generate value-add landing pages at scale.
6. "Clay for content" repurposes content across every use case, vertical, and buyer profile to maximize coverage in the prompts buyers actually search.
7. Reddit marketing agents that find high-intent conversations, draft on-brand responses, and post automatically. Self-improving based on editorial feedback over time.
8. Newsletter agent that produces Morning Brew-style curation newsletter. Nurtures leads with only a few minutes of human work required.
9. Account based marketing (ABM) agent that monitors target accounts and tells your team exactly when to reach out and what to say.
10. Custom deck and proposal generator with interactive interfaces where leads mix and match investment scenarios and run portfolio calculations in real time.
11. Earnings analyst agents that monitor calls in real time and delivers briefs with citations.
12. AI product lines for services firms. Custom AI software built for agencies and consultancies to deploy with their own clients. They stop renting tools and start owning IP.
Here are 12 agents and AI tools we've built for clients:
1. AIEO visibility platform that monitors brand presence across LLMs, generates high intent prompts to track, and delivers execution plans.
2. Sales call roleplay trainers where reps pitch voice AI prospects trained to talk like their real buyers. Personalized, post-call coaching slashed onboarding time and boosted win rate.
3. AI GTM agents that build prospect lists, research every account, and write personalized messaging. Creative campaigns that used to take days now takes minutes.
4. Lead scraping and enrichment agent that processes tens of thousands of leads, scores by problem symptoms, enriches contact info, and syncs to CRM. Sellers open their laptop to qualified leads every morning.
5. AIEO/SEO content agent that pulls customer insights, keyword research, and proprietary data to generate value-add landing pages at scale.
6. "Clay for content" repurposes content across every use case, vertical, and buyer profile to maximize coverage in the prompts buyers actually search.
7. Reddit marketing agents that find high-intent conversations, draft on-brand responses, and post automatically. Self-improving based on editorial feedback over time.
8. Newsletter agent that produces Morning Brew-style curation newsletter. Nurtures leads with only a few minutes of human work required.
9. Account based marketing (ABM) agent that monitors target accounts and tells your team exactly when to reach out and what to say.
10. Custom deck and proposal generator with interactive interfaces where leads mix and match investment scenarios and run portfolio calculations in real time.
11. Earnings analyst agents that monitor calls in real time and delivers briefs with citations.
12. AI product lines for services firms. Custom AI software built for agencies and consultancies to deploy with their own clients. They stop renting tools and start owning IP.
Today we’re launching the OpenAI Deployment Company to help businesses build and deploy AI.
It's majority-owned and controlled by OpenAI. It brings together 19 leading investment firms, consultancies, and system integrators to help organizations deploy frontier AI to production for business impact. openai.com/index/openai-l…
16
16
340
80,148
Box CEO @levie built his 3,000-person company's AI strategy with 3 simple questions:
1. More information
"Imagine if that particular job had access to like 10X or 100X more information at its disposal, what would you do differently in that job function?"
" If you were a sales lly 10 or 100 times more intelligence about your accounts you're going after...how much better would you be able to deliver the right message or the right value proposition to that customer?"
2. Tool integration
"What parts of jobs are we spending an inordinate amount of time moving things between systems?"
"Now in 2026, it's not only accurate answers in a knowledge base, but performing tasks and actually those tasks are over time getting better and better because there's some kind of feedback loop and eval process and continuous improvement that's going on."
3. Surplus reinvestment
"If you're doing things right, hopefully now you have a surplus."
"What can you automate? Do that to the maximum degree, but then what do you now benefit from by having those extra surplus resources, and how do we go apply those?"
1. More information
"Imagine if that particular job had access to like 10X or 100X more information at its disposal, what would you do differently in that job function?"
" If you were a sales lly 10 or 100 times more intelligence about your accounts you're going after...how much better would you be able to deliver the right message or the right value proposition to that customer?"
2. Tool integration
"What parts of jobs are we spending an inordinate amount of time moving things between systems?"
"Now in 2026, it's not only accurate answers in a knowledge base, but performing tasks and actually those tasks are over time getting better and better because there's some kind of feedback loop and eval process and continuous improvement that's going on."
3. Surplus reinvestment
"If you're doing things right, hopefully now you have a surplus."
"What can you automate? Do that to the maximum degree, but then what do you now benefit from by having those extra surplus resources, and how do we go apply those?"
6
5
34
2,983
Box CEO Aaron Levie (@levie) drops the truth about enterprise AI adoption today.
AI isn't spreading evenly across jobs.
Adoption is highly concentrated in engineering because it's verifiable.
"We don't have the systems in place for verification...as easily or perfectly as "Your job is not really just producing digital information in a way that is like testable...you're negotiating in an interpersonal way."
Subjective job functions like marketing will require more human in the loop until we: 1) find creative verification methods like synthetic users and/or 2) reduce the need for verification by improving the quality of the first draft.
AI isn't spreading evenly across jobs.
Adoption is highly concentrated in engineering because it's verifiable.
"We don't have the systems in place for verification...as easily or perfectly as "Your job is not really just producing digital information in a way that is like testable...you're negotiating in an interpersonal way."
Subjective job functions like marketing will require more human in the loop until we: 1) find creative verification methods like synthetic users and/or 2) reduce the need for verification by improving the quality of the first draft.
9
6
36
3,566
It looks like there's huge demand for AI products and services. There is, but it's mostly coming from AI-native tech companies and large enterprises.
On Friday, I went to an event with 10 main street business owners (bookkeeping, house cleaning, business broker, etc.).
Their rwhelmingly negative. Most think AI still hallucinates all the time.
@chamath's read on Anthropic's S-1 on the latest All In pod shows the same gap:
"They have barely scratched the surface of the commercial working world."
"There's about 40 million individuals and small businesses that are spending under $100k."
"There's about 1,650 that spend between one and 10 million, and only 250 customers spend 10 million and more."
"A big portion of those are going to be AI natives, folks that are venture funded. So, you have to strip those folks away. And then what's left over is probably a couple of hundred global 2000 type companies."
The demand for AI is huge, but it's concentrated.
The next big wave of revenue is the mainstream market.
Unlocking will take a lot of education and marketing. And it's needed if Anthropic wants to grow into a $2 trillion valuation.
On Friday, I went to an event with 10 main street business owners (bookkeeping, house cleaning, business broker, etc.).
Their rwhelmingly negative. Most think AI still hallucinates all the time.
@chamath's read on Anthropic's S-1 on the latest All In pod shows the same gap:
"They have barely scratched the surface of the commercial working world."
"There's about 40 million individuals and small businesses that are spending under $100k."
"There's about 1,650 that spend between one and 10 million, and only 250 customers spend 10 million and more."
"A big portion of those are going to be AI natives, folks that are venture funded. So, you have to strip those folks away. And then what's left over is probably a couple of hundred global 2000 type companies."
The demand for AI is huge, but it's concentrated.
The next big wave of revenue is the mainstream market.
Unlocking will take a lot of education and marketing. And it's needed if Anthropic wants to grow into a $2 trillion valuation.
16
0
29
3,338
"100% of code written by Vercel engineers was written by agents." - Vercel CTO Malte Uble (@cramforce)
So are engineers cooked?
Malte says AI creates more demand for engineers, not less.
"You would think, 'well, it's so much easier to make code now, there must be a engineers.' And in fact, the opposite is happening. It's going up."
Demand is growing because more software is being built:
1. More companies are choosing build over buy
2. Agents are a new kind of software that open up new use cases
So are engineers cooked?
Malte says AI creates more demand for engineers, not less.
"You would think, 'well, it's so much easier to make code now, there must be a engineers.' And in fact, the opposite is happening. It's going up."
Demand is growing because more software is being built:
1. More companies are choosing build over buy
2. Agents are a new kind of software that open up new use cases
2
3
7
528
The Proton CTO's AI Manifesto is a literal time portal. If you want to see your future, read it:
"Your job has fundamentally changed. It is no longer [job]...Your job is to help the LLM [do your job]."
"AI needs to be be able to work autonomously and continuously with less and less human babysitting."
"Your job has fundamentally changed. It is no longer [job]...Your job is to help the LLM [do your job]."
"AI needs to be be able to work autonomously and continuously with less and less human babysitting."
Here is Proton CTO's AI manifesto, that I obtained multiple copies of 404media.co/manifestos-fea…
6
1
9
998
More companies are choosing to build their own tools than ever before.
But the number one fear holding more companies back from building is MAINTENANCE.
We decided to squash that fear and enable more companies to harness the full power of AI in their business.
My company Atherial now provides Commission-Only Maintenance.
We align on your desired outcome before the build.
After the build, we only get paid when you make money or save money from it.
In return, we handle model upgrades, fixes, and optimizations.
Maintenance is a real responsibility. Railway's recent hack showed why it's so important.
But it shouldn't block companies from taking advantage of the incredible models we have access to today.
And now it doesn't.
Let's build something great together.
But the number one fear holding more companies back from building is MAINTENANCE.
We decided to squash that fear and enable more companies to harness the full power of AI in their business.
My company Atherial now provides Commission-Only Maintenance.
We align on your desired outcome before the build.
After the build, we only get paid when you make money or save money from it.
In return, we handle model upgrades, fixes, and optimizations.
Maintenance is a real responsibility. Railway's recent hack showed why it's so important.
But it shouldn't block companies from taking advantage of the incredible models we have access to today.
And now it doesn't.
Let's build something great together.
0
0
8
749
Data broker = micro1 etc or be an affiliate
Vertical AI = harvey, photo editing, etc
AI services = FDE, consulting etc
Boring biz = marketing agency, tooth paste company, etc and run AI behind the scenes
Vertical AI = harvey, photo editing, etc
AI services = FDE, consulting etc
Boring biz = marketing agency, tooth paste company, etc and run AI behind the scenes
0
0
1
104
My buddy Brett built AI that dissected 4,310 ads by Fortune 500 brands.
It broke down every piece of creative into 5k+ features like hooks, visuals, audio, on-screen text and timing.
Here's one surprising tactic that's working on video platforms like IG and YT right now: e with neutral facial expressions gets more engagement than people with smiles.
Why? My hypothesis:
People crave truth. Smiles look fake.
My favorite content creators share their real life stories, even when they're dark.
I'm all for providing insight and value in content, but be a real person with a personality and a story too.
It broke down every piece of creative into 5k+ features like hooks, visuals, audio, on-screen text and timing.
Here's one surprising tactic that's working on video platforms like IG and YT right now: e with neutral facial expressions gets more engagement than people with smiles.
Why? My hypothesis:
People crave truth. Smiles look fake.
My favorite content creators share their real life stories, even when they're dark.
I'm all for providing insight and value in content, but be a real person with a personality and a story too.
2
0
2
355
Check out the full report with 12 actionable insights
collinearity.ai/research_repor…
collinearity.ai/research_repor…
0
0
0
105
There’s a tragedy of the commons problem here.
If one expert sells their sauce to AI companies, it's nbd.
But if everyone does it, literally all wealth accrues to Sam and Dario.
If one expert sells their sauce to AI companies, it's nbd.
But if everyone does it, literally all wealth accrues to Sam and Dario.
I’ve been surprised by how little humanity I’ve seen in startup media the past year
most of the public has no idea who Mercor is, but if they understood what they do they’d be furious.
experts get paid once to hand over skills that took a lifetime to cultivate for models to replicate their efforts forever, without ownership. it’s robbing from the collective. I have seen so few tweets saying anything honest here.
my gut reaction watching this is worry. I am worried. I see through the words and the pitch. being an investor great, being an employee, exciting. but who’s at the wheel on ethics? for any of these companies.
“someone’s going to do it”, but who’s the authority on ethics in this industry. it seems like no one.
1
0
3
532
The most AI-native people I know have done 0-1 startups.
They know how to replace themselves:
> start a job function or channel
> figure out how to get it working
> write up the playbook
> hire someone to execute and optimize
> onboard and manage them
Now you just build and train AI instead of hiring. Pretty much the same thing otherwise.
They know how to replace themselves:
> start a job function or channel
> figure out how to get it working
> write up the playbook
> hire someone to execute and optimize
> onboard and manage them
Now you just build and train AI instead of hiring. Pretty much the same thing otherwise.
17
0
24
1,138
Everyone says technical moats are dead, but it seems like to build a vertical AI company, you need $50M to spend on:
- expert evals from mercor
- company data from micro1
- tokens from claude code/cursor
- cracked ai engineers
Then you need distribution. So you spend more money on influencer marketing, ads, organic content, and cold email.
"Anyone can build anything" is a lie.
Anyone can vibe code SaaS.
Not everyone can build a winning AI company.
To do that you need a lot of capital.
I'd hate for the AI boom to be only a rich man's game.
What are the best ways to make money from the AI boom besides building vertical AI?
- expert evals from mercor
- company data from micro1
- tokens from claude code/cursor
- cracked ai engineers
Then you need distribution. So you spend more money on influencer marketing, ads, organic content, and cold email.
"Anyone can build anything" is a lie.
Anyone can vibe code SaaS.
Not everyone can build a winning AI company.
To do that you need a lot of capital.
I'd hate for the AI boom to be only a rich man's game.
What are the best ways to make money from the AI boom besides building vertical AI?
Ben Horowitz (@bhorowitz): The laws of physics of company building changed.
You can throw money at problems now.
"If you have the data and you have enough GPUs, you can solve damn near anything."
Technical moats aren't dead. They're just turning into capital moats.
data and compute can catch up (or stay ahead).
Strong teams that are well capitalized get exponentially more leverage than before.
35
3
30
3,817
micro1 CEO @aliansarinik breaks down the difference between writing evals and writing a rubric:
Writing the eval means an expert rates the outputs.
Writing the rubric means the expert describes a good outcome.
An agent can then verify its own work against the rubric forever. these questions and rating it and so forth, you get [experts] to actually describe with a bunch of different verifiers, what a good outcome would be."
"These are what's called binary rubric items which you can have a model grade another model with because you're objectively saying this thing should be included."
"No matter what the model's knowledge is in legal memos or legal broadly it can just check if that thing is there."
"The beautiful part is that you can rerun this...the human experts don't have to look at it again and rewrite anything. They already wrote the objective criteria."
When a new model ships, you rerun your agent on the same rubric and get a score to see if you should swap to that new model.
Same loop for improving your agent. Change one thing, rerun on the rubric, keep what scores higher. That's hill climbing.
Writing the eval means an expert rates the outputs.
Writing the rubric means the expert describes a good outcome.
An agent can then verify its own work against the rubric forever. these questions and rating it and so forth, you get [experts] to actually describe with a bunch of different verifiers, what a good outcome would be."
"These are what's called binary rubric items which you can have a model grade another model with because you're objectively saying this thing should be included."
"No matter what the model's knowledge is in legal memos or legal broadly it can just check if that thing is there."
"The beautiful part is that you can rerun this...the human experts don't have to look at it again and rewrite anything. They already wrote the objective criteria."
When a new model ships, you rerun your agent on the same rubric and get a score to see if you should swap to that new model.
Same loop for improving your agent. Change one thing, rerun on the rubric, keep what scores higher. That's hill climbing.
Databricks Chief AI Scientist Jonathan Frankle (@jefrankle) says the bottleneck to enterprise AI adoption is evals, not models.
"The models are good enough at a lot of really complex enterprise tasks and knowledge tasks"
"the real challenge is describing what you want AI to do did it"
5
5
102
8,164