Rohan Punamia

if non-LLM development tech companies are laying tech people off and blaming LLMs while the LLM companies are hiring tech people like mad, the simplest explanation for me is that non-LLM companies are probably using it as an excuse. laying people off SUCKS... its hard emotionally for most people and its always nice when you can blame it on something other than your own mistakes such as: a).overhiring b) building the wrong thing c) the randomness of the business cycle d) short term profitability vs long term employee value/loyalty (for public companies a big one) etc.

14

1 Comment

I talked with Jensen Huang about AGI. Here's the real truth about engineering: 45 minutes to process 200 deals. 𝐋𝐚𝐦𝐛𝐝𝐚 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐬 𝐭𝐢𝐦𝐢𝐧𝐠 𝐨𝐮𝐭 𝐞𝐯𝐞𝐫𝐲𝐰𝐡𝐞𝐫𝐞. The audit system pulled deal data from HubSpot, notes, emails, calls, ran each through an LLM for extraction, then enriched the output. One deal took 12-15 seconds. Multiply by 200 sequentially and the pipeline couldn't survive its own duration. 𝐈 𝐜𝐡𝐨𝐨𝐬𝐞 𝐭𝐡𝐞 𝐟𝐚𝐧-𝐨𝐮𝐭 𝐩𝐚𝐭𝐭𝐞𝐫𝐧. Fan-out splits workload across parallel workers instead of processing items one after another. One orchestrator divides the work, multiple workers execute simultaneously, results merge back. Restaurant kitchen logic, one chef cooking 50 orders means cold food, five chefs with an expeditor means everyone eats on time. 𝐄𝐂𝐒 𝐫𝐮𝐧𝐬 𝐭𝐡𝐞 𝐨𝐫𝐜𝐡𝐞𝐬𝐭𝐫𝐚𝐭𝐨𝐫. 𝐋𝐚𝐦𝐛𝐝𝐚 𝐡𝐚𝐧𝐝𝐥𝐞𝐬 𝐝𝐞𝐚𝐥 𝐩𝐫𝐨𝐜𝐞𝐬𝐬𝐢𝐧𝐠 𝐢𝐧 𝐛𝐚𝐭𝐜𝐡𝐞𝐬 𝐨𝐟 𝟏𝟓. 𝐅𝐢𝐫𝐬𝐭 𝐯𝐞𝐫𝐬𝐢𝐨𝐧 𝐟𝐚𝐢𝐥𝐞𝐝 𝐢𝐦𝐦𝐞𝐝𝐢𝐚𝐭𝐞𝐥𝐲. Redis couldn't handle the payloads. Lambda's 6MB limit plus memory pressure corrupted bundles. Moved everything to S3 with gzip compression. Then 14 Lambdas finished within milliseconds of each other. Each one asked: "Am I last? Should I trigger collection?" Race condition. Redis SETNX solved it — first Lambda sets a lock key, everyone else exits gracefully. Static timeouts failed unpredictably. 𝐒𝐰𝐢𝐭𝐜𝐡𝐞𝐝 𝐭𝐨 𝐝𝐲𝐧𝐚𝐦𝐢𝐜: 𝐭𝐢𝐦𝐞𝐨𝐮𝐭 = 𝐛𝐚𝐬𝐞 + (𝐝𝐞𝐚𝐥𝐬 × 𝐬𝐞𝐜𝐨𝐧𝐝𝐬_𝐩𝐞𝐫_𝐝𝐞𝐚𝐥). Batch size took tuning. Too small and invocation overhead killed throughput. Too large and HubSpot rate limits killed everything. Landed on 15 deals per Lambda. 200 deals. 3 minutes. No timeouts. What's the dumbest bug that cost you the most time? 😆

17

The models VCs rely on for comp and headcount just don’t fit anymore. Here's what I'm seeing as a software founder myself: The traditional playbook for headcount growth is shifting. Instead of the linear scaling we used to see, where revenue growth directly correlated with team expansion, companies are hiring fewer people. And those they do hire? They're often commanding significantly higher compensation. This creates a real problem for the benchmarking tools we all use (Carta, Pave, salary.com). These platforms often use 3-4 year old data in their aggregated benchmarks. And in a market moving this fast, that historical data might not reflect current reality. And even though the benchmarking companies will say they’re “accounting for it,” who really knows? There isn’t much transparency around which data points are most recent, and for a lot of new roles (like AI orchestration ops or forward-deployed engineers), there just isn’t enough data to make a statistically sound recommendation anyway. So when a portfolio company asks, “How many engineers should we hire next quarter?” or “How much should we offer our forward-deployed engineer?” It feels like there is a big gap in real data for VC talent leaders to make an informed answer. On top of it all, there's the remote work complexity... Say you want to hire an exceptional engineer living in Portugal. The benchmarking tools will tell you what the average Portuguese engineer makes. But that's comparing apples to oranges. An engineer with a Silicon Valley startup skill-set who happens to live in Portugal isn't competing with local telecom companies. They're competing globally. (thanks to Rippling / Deel) Offer them $50K because "that's the Portuguese market rate"? It won't fly. So, how do we create fair compensation frameworks when people work across different countries with vastly different costs of living but similar competitive landscapes? No one has fully solved this yet. We're all learning together. And right now, one of the best resources VCs and talent leaders have is each other, comparing what’s actually happening across portfolios. Ultimately, that’s how the market gets made. What are you seeing in your portfolio or at your startup? Are the traditional comp benchmarking tools still serving you well, or are you looking for one-off examples?

12

2 Comments

📝How much equity should your first engineering hire get? 🐎Sure 1% is “what they all say” but here’s the data from the horses mouth aka real cap tables from Carta. 🧍‍♂️My take? 75% percentile is 4%+ most likely because the first engineerinf hire is filling a technical co-founder gap. 🔮As engineering teams continue to get leaner and more augmented, will this comp go up, or will it hold steady with founders keeping more for themselves? 🥑Thx to Peter Walker for the original data post (and for buying my avocado toast this morning).

116

18 Comments

The SaaSpocalypse Isn't Real. But the Opportunity for FMs Is. Wall Street panicked this month. The S&P 500 Software Index dropped 13% in a single day. OpenAI launched a platform that wants to be the "operating system of the enterprise." Investors started talking about the death of SaaS. They're calling it the SaaSpocalypse. Marc Benioff addressed it head-on during Salesforce's earnings call this week. He said the word six times. Then joked about a SaaSquatch eating it. Classic Benioff. But buried under the Wall Street drama is something that actually matters for facilities teams. Here's the debate in plain English: OpenAI says: AI agents should sit on top of everything. Your business software — your CRM, your CMMS — is just plumbing feeding the AI. Salesforce says: Agents are only as good as the data they sit on. Without governance, security, and audit trails, your AI agent is just a really expensive intern making stuff up with no supervision. I know where I land. Think about what happens when an AI agent creates a work order. Where does it go? Who approved it? What's the audit trail? If the agent assigns a vendor, does it know the vendor's insurance expired last month? That's what a system of record does. It's not glamorous. It's data governance. It's making sure that when OSHA shows up, you can pull the inspection history and prove you did the work. You can't shove AI at chaos and expect it to make things better. You need clean data first. Defined processes. A system that knows who's accountable for what. The SaaSpocalypse hasn't shown up in the numbers. Salesforce just reported $11.2 billion in quarterly revenue, up 12%. Their AI and data products — Agentforce and Data 360 — generate over $2.9 billion in annual recurring revenue, up over 200% year-over-year (boosted by the Informatica acquisition, but the organic growth is still over 100%). SaaS isn't dying. It's becoming the foundation agents need to function. Here's what this means for you. IFMA estimates 40% of existing facilities managers will retire this year. There's a projected shortfall of 53% in open positions. As many as 158,000 unfilled FM roles through the end of the decade. AI agents aren't replacing you. They're the only way you're going to manage the workload that's coming. But they need clean data. Governed systems. Proper workflows. A system of record. The facilities teams that get ahead of this — the ones who clean up their asset registers, who move from spreadsheets to proper platforms, who learn how these tools work — those are the teams that will thrive. The FMs who say "that's IT's problem" are the ones who'll get left behind. Not by AI. By their peers who decided to lean in. Don't be afraid. Get curious. The industry needs you more than ever. #FacilitiesManagement #AI #CMMS #IFMA #SaaS #FMJobs #DataGovernance

9

The $10M ARR myth that nobody wants to talk about → Stripe just dropped a bomb: Double the startups hit $10M ARR in 3 months in 2025 vs 2024. Everyone's celebrating. I'm concerned. Here's why this "achievement" might be the worst thing happening to the startup ecosystem: → Speed doesn't equal substance → Revenue doesn't mean profitability → Fast growth often masks fundamental problems The uncomfortable truth? Most of these "instant success" stories will be dead in 18 months. Why? Because hitting $10M ARR in 90 days creates a dangerous illusion: • You think you've figured out product-market fit (you haven't) • You scale before validating retention (customers churn) • You raise capital on vanity metrics (burn rate explodes) • You hire aggressively without culture foundation (team collapses) The data Stripe conveniently left out? Customer retention rates. Churn metrics. Actual profitability. I've seen this movie before. The 2021 "growth at all costs" era destroyed more companies than it created unicorns. AI is making it easier to acquire customers fast. But AI can't fix: - Poor unit economics - Weak value propositions - Non-existent customer support - Unsustainable business models The real question isn't "How fast can you hit $10M ARR?" It's "Can you still be in business 3 years after hitting $10M ARR?" Because sustainable growth > explosive growth. Every. Single. Time. Are we building the next generation of category-defining companies, or just creating a new bubble of revenue-rich, profit-poor startups destined to implode? What's your take – is ultra-fast ARR growth a feature or a bug of the AI era?

5

Build in public vs stealth mode When stealth wins? When does a public building help? And which founders can actually raise big money quietly? In just the past six months: Periodic Labs came out of stealth with a $300M seed, Eridu emerged with over $200M in funding, Simile surfaced with $100M, Gradium launched after three months in stealth mode with a $70M seed round. A lot of startup advice today says:   - Build in public. - Post the journey. - Share the roadmap. - Turn progress into content.   That advice is useful.   For many companies, visibility is a growth tool.   It helps you learn faster. It helps you find users. It helps you attract early believers.   But stealth mode is just a different strategy for a different kind of company.   Here is the real split:   1. Build in public works when feedback is an advantage   - consumer products - community-led tools - creator products - many SaaS companies   In these businesses, the biggest risk is invisibility. - You need users. - You need reactions. - You need signals from the market. - You need to know quickly what people care about and what they ignore.   Public buildings help shorten that loop.   2. Stealth mode works when secrecy is the advantage   - AI infrastructure - chips - biotech - robotics - defense - security - hard enterprise systems   In these markets, announcing too early can create real downside. - You can get copied before the moat is real. - You can attract pressure before the product is ready. - You can expose your hiring targets.   3. Big stealth rounds do not go to random founders with an idea   This is where the conversation usually gets blurry.   Not every founder can raise a huge round in stealth.   Investors back founders with one of three forms of trust:   - Repeat founders who already built, scaled, or exited a company. - Researchers or engineers building something very few people can build. - Insiders who know the buyers, the problem, and how budgets move.   Investors are funding credibility behind the silence.   4. Stealth does not mean there is no traction A stealth startup may look empty in public. But in private, it may already have: - design partners - LOIs - pilot customers - top hires - patents - technical milestones - regulatory progress So while the outside world sees a blank page, investors see proof. 5. Build in public and stealthily solve different problems Built-in public helps when you need: - attention - feedback - distribution - community - trust through transparency Stealth helps when you need: - time - focus - protection - controlled narrative - quiet recruiting Neither approach is automatically better. Each works only when matched to the real risk. That is why the real question is not: Should I build in public or stealth? It is: What creates a stronger advantage for this company right now? What did I miss? ♻️ Repost to help founders in your network. Follow Leon Eisen, PhD for more VC insights.

114

46 Comments

$100 billion of "Neo Labs" are quietly breaking the VC model... Everyone on LinkedIn keeps talking about the traditional math of venture capital: Power laws, fund sizes, IPO paths, and portfolio construction. That math works when capital is deployed into companies. Entities built with clear products, defined GTM paths, and a realistic shot at commercial liquidity. But we’ve entered the era of the Neo Lab. These are VC-funded AI institutes that operate more like academic research centers than startups. They prioritize "Scientific Superintelligence" over quarterly earnings. The scale of this shift is staggering: Safe Superintelligence Inc.(SSI): Raised $3B at a $32B valuation to chase a single technical milestone. Thinking Machines Lab: Mira Murati’s new lab, which hit a $12B valuation (and is reportedly eyeing $60B) on the strength of research pedigree alone. Isomorphic Labs: Alphabet’s spin-off recently took its first $600M in external funding to "solve" biology. Lila Sciences: Building "AI Science Factories" with $550M to automate the scientific method. Black Forest Labs: A $3.25B powerhouse for visual intelligence that prioritizes open research over enterprise SaaS sales. Where the model breaks: Venture capital expects asymmetry. Historically, that asymmetry comes from distribution, pricing power, and execution. It does not come from papers, demos, or benchmarks alone. One of these has to hit a multi-trillion dollar valuation to make VC math work. Are we witnessing the most expensive science experiments in history, a bubble, or a fundamental evolution of how humanity innovates? 

10

1 Comment

Secondary sales were 3% of VC exits in 2015. Today, they’re 31%. Liquidity is decoupling from company maturity. Returns depend less on timing the IPO…and more on managing ownership strategy from Series B onward. It’s a large shift for the VC community. Great report from PitchBook here: https://lnkd.in/exKYdVAW Acrew Capital Angeles Investors Insight Partners The Tech Series Nasdaq

2

Why the traditional SF model is failing AI startups and NYC is winning. Silicon Valley burns $2M/month perfecting models, NYC founders close $100K deals with "inferior" tech. Customers don't care if your model is 2% more accurate. They care if it actually works in their workflow. NYC cracked the code. SF: "Our model outperforms GPT-5 on reasoning tasks!" NYC: "Goldman Sachs just signed a 3-year contract." Guess who's winning... NYC AI startups are hitting $1M ARR 40% faster than SF counterparts. Not because of better tech. Because of better sales execution. The Valley delusion is expensive: $500K engineers building perfect products that nobody wants to buy. NYC's unsexy approach is to ship fast, sell faster, iterate based on actual revenue. The next $10B AI company won't emerge from the city with the best researchers. It'll come from whoever understands customers and sales best. SF built the incredible infrastructure. NYC is capturing all the value. Distribution is the new moat. Game over.

26

12 Comments

With the launch of Claude Co-Work (and tools like it), I feel like every week in 2025 carries the weight of a decade. Some thoughts: 1. The technology is genuinely different this time. Claude Co-Work isn't an incremental improvement — it's a 100x leap over what existed 12 months ago. If you haven't used it yet, stop what you're doing. 2. Markets. Software and advisory businesses face an existential reset .Not decline. Reset. Consulting, legal, accounting, financial advisory — these industries aren't disappearing, but they're heading toward 80% fewer people doing the same work. KPMG just told auditors to cut costs by 50%. This is the beginning, not the exception. 3. Real-world industries aren't safe either — they're just next. Industrials, energy, consumer goods, travel all have a natural (physical) moat. But consumer brands being built today are doing it with 80-90% less headcount than their predecessors. Any company older than 6 months that isn't adapting is already behind. Jack Dorsey's Block cutting 50% of its workforce despite strong growth is a preview, not an outlier. 4. The new bottleneck is energy, not labor. Once labor becomes abundant and cheap, the constraint shifts. Whoever can produce energy at scale and reasonable cost will have the most important infrastructure asset of the next 20 years. 5. The job market is being structurally changed for good and this is not a minor adjustment. Bankers, lawyers, accountants, software engineers — some roles will survive the way painters survived the printing press. Most won't. The talent that moves fastest onto new tools will capture a disproportionate share of the upside. We're entering a period of massive productivity unlocks and equally massive collateral damage. The future is bright. The transition is not going to be comfortable. Go test out some of those new tools.

119

19 Comments

BREAKING: Palantir Technologies moves its HQ to Miami 🌴 Another Silicon Valley giant just voted with its feet. Palantir is relocating its headquarters from Denver to Miami, becoming the latest major tech player to anchor itself in South Florida. CEO Alex Karp has long criticized Silicon Valley culture, arguing the company increasingly “shares fewer of the technology sector’s values and commitments.” This is not random. Florida offers: • No state income tax • A business-friendly regulatory climate • An increasingly dense network of tech founders and billionaires Meanwhile, in California, there’s a proposed 5% wealth tax on residents worth over $1B. On Polymarket, there’s currently a 36% probability it passes in 2026. Capital is mobile. Talent is mobile. Headquarters are mobile. Palantir, founded in Palo Alto in 2003, already left California once for Denver in 2020. Now it’s doubling down on Miami, where co-founder Peter Thiel has an established presence. Zoom out: • ServiceNow expanding in West Palm Beach • Citadel operating from Miami • A wave of founders and executives setting up in South Florida The question isn’t whether Miami is becoming a serious tech hub. It’s whether Silicon Valley’s policy direction is accelerating its own decentralization. What’s next: Austin? Dubai? Singapore? Or is Miami quietly becoming the capital of the new American tech right?

45

11 Comments