Oli Nold

Workforce datasets are static. A snapshot. A list. A moment in time. But companies are not static. They grow. They contract. They shift role composition. They reallocate talent before revenue changes show up. So instead of building another database, I built a longitudinal company-year panel. ~2.5M normalized U.S. companies. ~387M company-year rows reconstructed from historical experience timelines. Median 7 years of workforce history per company. Not profiles. Not contact records. Company-year intelligence. For each company and each year: • Observed headcount • Growth rate • Role distribution shifts • Structured entity normalization The real asset isn’t volume. It’s the ability to ask: – When did this company actually start scaling? – Did engineering grow before sales? – How did workforce composition change pre-funding? – Which segments show consistent multi-year expansion patterns? Longitudinal structure turns raw records into signal. Investors call it alternative data. Strategists call it market intelligence. AI teams call it training infrastructure. I call it organizational time-series intelligence. Building this in public. #infrastructure #database #patternrecognition

Everyone jokes about March Madness killing productivity. Here's what the workforce data actually shows: Companies that go dark on hiring during cultural events (major sports, holidays, summer) consistently underperform their hiring targets by 23% annually. The ones that hire straight through? They're the signal in the noise. When your competitor pauses because "no one's paying attention anyway," that's exactly when serious operators move. We've tracked this pattern across 250M+ employment records over 15 years. The real productivity loss isn't employees streaming games. It's leadership teams that mistake their own distraction for market reality. Your best candidates are still looking. Your fastest growing competitors are still hiring. The bracket can wait.

AI fraud is scaling fast. Fake resumes. Fake employees. Entire fake work histories generated in seconds. But here's the thing: AI-generated profiles are too clean. Real careers are messy. Skills evolve. People move between companies that actually existed at the time. Their progression makes sense when you map it against headcount data and industry shifts. Fake profiles? They're perfect. And that's the tell. When you're looking at millions of workforce records, the anomalies jump out. Credentials that appear across 50 "different" candidates. Employment histories at companies that never hired during those periods. Zero skill migration patterns. The fraud scales, but so does the detection surface. Most companies are still doing manual resume reviews. One at a time. That worked when fraud was human-scale. Now? You need to see the full picture. Cross-reference against actual workforce movement data. Spot the patterns that only show up at scale. After 12 years building data businesses, I'll tell you this: the companies that win the talent war won't be the ones with the best job posts. They'll be the ones who can separate signal from noise when everyone else is drowning in generated garbage.

Everyone's debating whether AI will bankrupt the grid. Meanwhile, our data shows specialized data center roles (cooling engineers, energy optimization specialists, sustainability architects) grew 340% at tech companies between 2020-2024. Here's what that tells you: the companies building AI infrastructure already know energy is the constraint. They've been hiring for it aggressively for four years while the rest of us argued about whether it's a problem. The real tell? These roles are concentrated at three types of companies: hyperscalers building their own data centers, chip designers who need to prove efficiency, and a handful of energy tech startups that just got very interesting acquisition targets. CMU's research is smart, but the talent migration happened first. It always does. Workforce data is a leading indicator, not a lagging one. The infrastructure play isn't just about who makes better chips. It's about who can actually staff the systems that keep those chips from melting.

Today we're introducing the world's first AI CMO. Enter your website and it deploys a team of agents to help you get traffic and users. Try it now at okara.ai/cmo

🚨 Urgent: Iranian drones in Dubai - 16.03 UPDATE! What are haters gonna say now haha?

@Sajwani Keeping the fake up as long as possible. It was always like that, it will always be like that. Normal behavior shortly before the total collapse

To all those residents making beautiful videos, taking pictures, and spreading positivity about Dubai and defending the UAE 🇦🇪 … THANK YOU 🤍 we will never forget you

Vibe coding AI labor exposure is cute. But here's what actually matters: which companies are hiring AI engineers while cutting marketing teams. Which skills migrated from finance to tech in 18 months. Where comp premiums actually showed up, not where McKinsey said they would. The highest-paid roles today weren't in our taxonomy in 2015. That's not a forecast. That's disruption already in motion. Everyone wants to predict AI impact. I track what already happened. 46.6M professional records. 15 years of workforce shifts. The patterns are brutal and clear. The jobs getting replaced aren't always the ones people think. And the new ones don't wait for your degree program to catch up.

@Fahadnaimb Dubai is as transparent as your posts. They fit very well.. matches the overall picture.

Emirates update: after temporary suspension/diversions, they’re resuming a limited schedule at Dubai Airport from 10:00 local time today. Some flights canceled due to the ongoing drone-related airspace issues.... re-accommodation offered to affected passengers. IMO: it’s frustrating for travelers, but Emirates is handling it transparently and trying to get things back on track as fast as possible

For the past few years we’ve been quietly building something unusual. A structured workforce dataset covering millions of companies. Hundreds of millions of professional records were normalized into company-year workforce observations: headcount growth, role distributions, organizational structure. What started as raw records slowly turned into a historical workforce intelligence archive. One of the first things we looked at was panel continuity. Across ~4M companies, about 64% appear consistently between 2018–2020. That level of continuity makes it possible to analyze how organizations actually evolve over time. Not job postings. Not press releases. Real workforce structure. You start seeing patterns: Engineering hiring spikes often appear months before companies scale. Sales expansion usually comes later. The dataset was originally built for internal analysis, but the more we explore it, the more interesting the signals become. Curious what others would analyze with a longitudinal workforce dataset like this.

7 million job openings…..Most of them are theater. Companies post roles they never seriously intend to fill. It keeps the growth story alive, signals expansion to investors, and reassures employees that the company is “scaling.” But the workforce data tells a different story. When you track actual headcount over time, many of these companies show almost no real growth. Same job postings circulating for months. Same roles reposted again and again. No meaningful increase in staff. Job boards measure ambition. Headcount timelines measure reality. Anyone can post a job. But when a company actually adds 40 engineers in a year, you see it immediately in the workforce sequence. company → year → headcount → role composition That is not marketing. That is budget. If you want to understand where companies are truly investing, stop watching job postings. Track who actually gets hired.

Most ML datasets about companies are built as snapshots. One row per company. One moment in time. That design already limits what a model can learn. Companies are temporal systems. They evolve through hiring, restructuring, and team composition changes. If your dataset ignores time, your model can only learn static attributes. While validating our workforce dataset, we learned something important about ML training data. Density beats length. The strongest windows are: 2018–2020 with about 2.7M companies 2016–2020 with about 2.3M companies 2015–2020 with about 2.1M companies For ML, a dense 3 to 5 year panel often produces stronger signals than longer but sparse timelines. Instead of a static company record, the model sees a sequence: company - year - headcount - role distribution - growth Once you structure the data like this, the model stops learning what companies are. It starts learning how companies change.

Starbucks just had their HR platform breached. Everyone's talking about PII exposure. They're missing the bigger picture. Attackers now have the org chart. Headcount by department. Compensation bands. Role hierarchies. Who reports to who. That's not just sensitive data. That's organizational intelligence. Here's what most people don't get: hedge funds pay six figures for this exact data. Legally. From third parties tracking workforce changes. Why? Because headcount moves predict revenue. Hiring patterns predict expansion. Compensation shifts predict retention risk. The irony: companies treat this data like it's locked in a vault, but it leaks constantly. LinkedIn updates. Job postings. Employee movement. Smart investors stopped waiting for breaches. They built infrastructure to track it systematically across millions of companies. The breach isn't the story. The value of workforce data is.

@ProductHunt @ycombinator @aaron_epstein Launch platforms are becoming real-time market intelligence layers, not just distribution channels. The real value emerges when launch data, user behavior, and founder outcomes are captured longitudinally to map how products actually gain traction.

We're teaming up with @ycombinator to get builders to launch. Schedule your launch for tomorrow, tag "YC application." and @aaron_epstein will review launches. Top ones could get a YC interview + potential funding. 👇

@dickiebush Consistency is really about minimizing recovery time after deviation, which is essentially a feedback-loop problem. The same principle shows up in data systems: organizations that measure and learn from every cycle compound performance much faster over time.

My personal definition of consistency: How quickly you get back on track doing things you’ve committed to doing. It’s not “doing something every day.” It’s about never falling off those things for an extended period of time.

Most people optimize for their own moment of friction, but systems thinking starts with understanding the full context of everyone involved. The same principle applies in data infrastructure: the best insights come from capturing the complete system over time, not isolated signals.

On a flight and there’s a kid who’s been screaming for the last 15 minutes. Lots of passengers making faces at the parents. Before you do that, just remember, the parents are having a 100x worse time than you are. Guaranteed. Put in some headphones and relax.

@patio11 Communication patterns like WTBU are essentially lightweight data protocols for reducing coordination errors in complex systems. Organizations that instrument and analyze these interaction patterns over time can turn them into a measurable layer of operational intelligence.

If this flavor text works for your org, wonderful, but find some equivalent flavor text which works for your org.

@levelsio The debate isn’t MCP vs APIs, it’s about which interfaces generate structured data that systems can learn from over time. The real leverage in AI infrastructure comes from the telemetry layer that captures how agents actually interact with tools and workflows.

Thank god MCP is dead Just as useless of an idea as LLMs.txt was It's all dumb abstractions that AI doesn't need because AI's are as smart as humans so they can just use what was already there which is APIs

Startup ecosystems compound fastest when knowledge transfer is paired with structured data about how companies are actually built over time. The next frontier is turning founder journeys, decisions, and outcomes into longitudinal datasets that help the next generation build with evidence instead of folklore.

Startup School is coming to India! 🇮🇳 Hear from founders like @harshilmathur of Razorpay, @viditaatrey of Meesho, @lkeshre of Groww, @mukundjha of Emergent and more. And join the best builders and hackers from across the country for a day of talks and sessions with YC partners.

The real moat in solving hard real-world problems isn’t just hardware or software, but the data generated from operating those systems over time. Companies that capture and structure that operational data build a longitudinal intelligence layer that compounds faster than the product itself.

“Heat-seeking missiles for pain. That is the perfect definition of people who really succeed at startups.” @Keller is the CEO and co-founder of @Zipline, the world's largest commercial autonomous delivery system. Back in 2011, he set out to solve the un-fancy, real-world problem of transporting life-saving medicines to people who need it. To do it, he’s enlisted builders who are willing to “crack some eggs.” On In Depth, he sits down with @brettberson to share why building hardware is 10X harder than you’d think. He shares: -Why Zipline had a 1% chance of working -The blood drone delivery launch in Rwanda that was near-catastrophic -What it took to scale to 5,000 hospitals -How he hires at Zipline, from execs who aren't above "plunging the toilets" to teenagers who’ve built submarines and robots in their garage -Why the most important companies of the next 10 years will be hardware

The most valuable companies in this cycle won’t just ship products faster; they’ll institutionalize decision-making through data. Teams that systematically capture execution data over time build a compounding intelligence layer that turns organizational learning into a durable competitive advantage.

"Ami lives in the future. Yinon is very present. Roy is an execution machine. And Assaf decides which voice leads the way." 📺 Watch @Shardul_Shah talk about the team behind the largest venture-backed acquisition in history on @TechCrunch @EquityPod: techcrunch.com/video/the-32b-…