What investors talked about on X, Oct 3 to 10, 2026

Topic 01·Fundraising

The fundraising advice worth saving

This was the most practical conversation of the week. Most of it came from a few people, mainly long threads by Elizabeth Yin and Steph Nass, so read it as advice from a handful of investors, not a consensus.

• Talk to customers before you pitch. Nicole DeTommaso (@nic_detommaso) said the main thing she looks for at pre-seed is proof you've had dozens of real customer conversations, with notes to back it up.
• Lead with a fact, not a slogan. Lolita Taub (@lolitataub) compared a "big market" line with a concrete line about paying customers. The concrete one wins.
• Make sure your email actually arrives. Steph Nass (@stephnass) pointed out that many cold emails to VCs land in spam. Send from your company's domain, set it up properly, and send in small batches. She also said short emails beat long ones.
• Pitch enough people. Elizabeth Yin (@dunkhippo33) revisited her "5-100-500" rule of thumb: about 100 investor meetings over 5 weeks to close a $500K seed. Her point was that 20 or 30 meetings isn't enough to tell whether a round can come together.
• A real look at a fund's inbox. Elizabeth also shared Hustle Fund's September numbers: 1,922 pitches, and about 70% of those companies were asking for valuations under $10M.
• Your closing questions count. Harry Stebbings (@HarryStebbings) said the questions a founder asks at the end of a meeting tell him a lot.

If you want to join in: Share one thing you learned the hard way in investor conversations, written for other founders. Don't attach an ask.

Topic 02·Venture funds

Big funds are getting bigger, and the early stage feels thinner

Several investors said the venture market is pulling apart. USV announced $900M in new funds (Nick Grossman, @nickgrossman; Fred Wilson, @fredwilson). Turner Novak (@TurnerNovak) highlighted a point from another investor's thread: some single mega-funds raised this year are bigger than all first-time funds combined, which leaves few firms doing true early-stage work.

Others questioned the math behind it. Alex Iskold (@alexiskold) said the outcome sizes his firm used to plan around no longer make sense. A Harry Stebbings clip with Venky Ganesan asked whether tranched rounds have become a habit instead of a signal, while Elizabeth Yin explained why she still likes them for small, early rounds. She also described a "tale of two cities": a few companies raise huge rounds fast, and everyone else still has to grind.

If you want to join in: Share what round structures you're actually seeing at your stage (tranches, check sizes, who's leading) and ask if others see the same.

Topic 03·AI agents

AI agents went mainstream, and Amazon pushed back

This was the biggest conversation of the week, with 23 investors posting about it. Amazon shutting off access for consumer shopping agents started it. Olivia Moore (@omooretweets) called the move both rational and a mistake, given how much of Amazon's profit comes from ads. Paul Graham (@paulg) saw it as an opening for a new competitor.

The bigger questions were about how agents will work. Andrew Chen (@andrewchen) asked whether personal agents can ever have network effects or will stay single-player tools. Nikita Bier (@nikitabier) said the web needs a standard way for agents to identify themselves. Seema Amble (@seema_amble) pointed to open questions about agents paying for things. Alex Iskold added a reality check: agents still need close human supervision on real tasks.

If you want to join in: Post one specific place where agents still break in your market, like a site that blocks them, a step that needs a human, or a payment that fails. Then ask what others see.

Topic 04·Software & moats

If AI can rebuild any software, what's still worth paying for?

A lot of investors argued that code itself is losing value. Naval Ravikant (@naval) said the models themselves may be the last real moat. Matthew Berman (@MatthewBerman) predicted software heads toward free, with value moving to things like compliance and security. Gale Wilkinson (@galeforcevc) noticed AI-native firms offering audit, tax and fund admin at about half the usual price.

There was pushback too. Angela Strange (@astrange) said deploying AI inside a company is as much about people and workflows as technology. Guy Wuollet (@guywuolletjr) asked where the network-effect businesses are. A Harry Stebbings clip with Dev Ittycheria argued that durable companies build a data loop competitors can't copy.

If you want to join in: Share what's genuinely hard to copy in your category (data, distribution, compliance, deployment) and invite pushback. Leave your product out of it.

Topic 05·Robotics & defense

Robots and defense: attention is moving to things that ship

Hard tech came up from a wide mix of investors. In a Turner Novak interview clip, Bilal Zuberi (@bznotes) said the most under-invested part of physical AI is actually getting robots deployed in the field. Jason Calacanis (@jason) argued for renting robots by the hour instead of selling them. Corinne Riley (@CorinneMRiley) poked fun at the weekly "GPT-3 moment for robotics" claims. Bilal also warned that hardware is still hard, and in some ways harder now that China supply chains are cut off.

Demand for private shares is pointing the same way. Turner Novak and Harry Stebbings both shared a Q3 secondaries list where defense names climbed alongside AI.

If you want to join in: Share a real lesson from the field about deployment, suppliers, or what customers actually adopt.

Topic 06·AI infrastructure

Compute is becoming a cost line, and data centers are getting political

The AI infrastructure conversation moved from "build more" to who pays and who objects. Ryan McEntush (@rmcentush) shared takes on Texas pausing data-center development and the strain on the grid. Guy Wuollet suggested that if inference becomes a company's biggest expense, it may need new kinds of financing, and he compared token budgets to how hedge funds hand out capital. Chamath Palihapitiya (@chamath) predicted a wave of consolidation in solar and batteries.

If you want to join in: If you see compute or energy costs firsthand, share one simple, honest data point and ask how others budget for it.

Quick hits

Also worth a look

• AI labs are buying bio data. Nikhil Raman (@uninsightful) said the big AI labs have become big, not-very-picky buyers of biological data, and that this will reshape the bio startup market.
• A few people pay for most consumer AI. a16z's new consumer AI report started a short conversation. Justine Moore (@venturetwins) highlighted the top 1% of spenders, and Alex Immerman (@aleximm) wrote about ads as the next revenue lever. Most posts came from one account, so treat it as one firm's view.