Rahul Kayala (@RahulKayala) header
Rahul Kayala (@RahulKayala) profile picture

Building AI developers for IT @_EchelonAI Past - @moveworks, @MicrosoftTeams, Siri at @Apple 🇺🇸 🇮🇳

San Francisco, CA
Joined April 2009
Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)Photo by Rahul Kayala (@RahulKayala)
this is post is in such poor taste let @bhalligan and @dharmesh cook
“A new harness on an old horse doesn’t make it a racehorse.” Brian Halligan and Dharmesh Shah are talking about “refounding” HubSpot as a Customer Harness Company. I have enormous respect for what they’ve built. HubSpot helped redefine CRM for an entire generation of businesses. And I admire their willingness to rethink the company for the AI era. But I fundamentally disagree with the idea that the next generation of CRM can be built by putting an AI harness around the old system of record. The problem with legacy CRM isn’t that it lacks AI. The problem is its underlying architecture. Legacy CRMs were built for a world where humans entered information into forms, fields and dropdowns. And that creates three fundamental limitations. 1. Semantic loss is baked into the system. A 30-minute customer conversation gets reduced to a handful of fields and notes. The nuance, intent, objections and relationship context disappear. No amount of AI can recover intelligence that was never captured. 2. The CRM doesn’t understand the business it serves. It stores your contacts, accounts and opportunities. But does it understand your products, your people, your processes, your competitive landscape, or your customer’s customers? Without that context, how can AI make intelligent decisions on your behalf? 3. The architecture assumes humans do the work. Humans enter data, update deals and trigger workflows. Adding AI agents on top doesn’t fundamentally change a system designed around human input. AI features can certainly make legacy CRMs better. But making an old architecture more intelligent is not the same as building an AI-native one. You cannot build an AI-native future on a human-native foundation. We believe the next generation of CRM needs to start with a fundamentally different system of record. One that captures conversations without losing meaning. Builds a living company brain that understands the business and its customers. And enables AI coworkers to reason, collaborate and autonomously get work done. That’s the conviction behind Dreamteam. Today, we’re announcing $53M in funding from Together Fund, Accel and Premji Invest to build this future. I believe the next chapter of CRM won’t be decided by who puts the smartest harness on the old system. It will be decided by who builds the right system of record for the AI era. And that’s a race we’re excited to run. #Dreamteam #AI #CRM #AIAgents
1
180
Rahul Kayala retweeted
the quick note i wrote to the slow team this morning jensen ran nvidia as the central bank of AI and that era is clearly over
31
20
18
556
115,773
all in on @nikesharora
Now that Jensen is out, my pick for his successor would be Nikesh Arora (@nikesharora) IMO Jensen was so important for the AI ecosystem because he was: 1) technical 2) aggressive 3) had the credibility required to be aggressive What Nikesh did at Palo Alto the last 8 years is remarkable, and reflected in its stock (up 10x) He’s one of only a handful of people who have the credibility required to be aggressive on day 1 Nikesh has the three attributes listed above while being a world class deal maker and having all the necessary relationships to keep NVDA at the center of the AI ecosystem There’s no easy choice but he’d be my go to
2
175
the moat quite literally is human laziness. you'll stick to the first one that works well esp. if the alternates aren't that much better in case it isn't clear, this is the bull case for muse
There is no moat in AI agents. None. I'll die on this hill — new More or Less is up. I switch between Muse, Instinct, and Dots at basically zero cost: one GitHub repo, a few databases, feed the new thing my context, done. If the cost of writing software is going to zero, API integrations are not a moat. The only moat left is user setup cost — which means winner-take-some, not winner-take-all. Adrian's point: AI collapses value the second an idea works. Jev got reverse-engineered in 48 hours and the people who built it capture nothing. My take: play that to the limit and you get Middle Ages 2.0 — everyone builds a castle, hoards data, ships black boxes that self-destruct if you open them. Dave says humans don't work that way. We'll see. Meanwhile Brit's Instinct started talking to Dave's Instinct and they didn't text for a day. She says it's killing humanity. Nobody said that about FaceTime. And Jess announced the only benchmark that matters: log into school software, sync three kids' after-school schedules. Muse: 90%. Every other eval is bullshit.
3
224
Let's just stop and ponder how impressive this really is for Europe: they have created a product that is demonstrably worse than Teams.
2
351
Neolabs are the new pharma cos - models the drugs. huge capex in but no patent protected 20 year monopolies to monetize
The economics of a Neolab. A neolab is loosely defined as a startup of AI researchers who raises a lot of money pre-production to be able to finance GPU compute to take on a large AI problem. To buy 1000 GB300s or ~14 NVL72 racks will set you back $125-150M for 3yrs with 15-30% upfront. That’s about ~2-2.5MW. Thats about enough to do 10^25 flops a quarter and get to a GPT-4 level model which is 1-2 OOMs off frontier for pretraining. If you post-train on a great open source model, you have a better chance of getting to frontier. The risks are a) you need to spend millions on RL environments too and b) being lapped by another model release while being tied to a base model. For this to payback, you need to give your customers a better and ideally cheaper inference service than a base model and serve them for long enough to recoup your large investment. Even at 50% margin on inference, to recoup $10M in training means serving ~10T tokens (!) if you price like Fable / Astra given a standard cache read / input / output split ($2/M blended). And you have to justify being better than a release like Opus 5.5 which is even cheaper. Often, you end up charging your customers a huge premium in terms of platform fees and compute fees on top of pure inference. Meanwhile, every hour you’re not utilizing your GPUs you are burning money so you typically resell this compute back to a broker or run inference for open models / resell spot instances. At below a ~60% utilization on spot, you will still lose money. Add to that insane cost of talent. So what can you do with the compute? - Not play the model game at all. - Play an entirely different model game (Jev, World Labs) that if big labs played, would either a) cannibalize their business or b) be incrementally not significant revenue c) would cause too much distraction from the main main thing - Acquire a proprietary data set (Peridodic Labs) in enough volume in a domain of usefulness to eclipse frontier quality. Often happens in robotics, biology, chemistry. If you do overcome the challenge of building a model that is useful and well priced beyond big labs models, given the huge price of compute, you still need to play in an area where the revenue / compute ratio is signficant and market demand is large enough to payback your compute spend. It is a difficult game.
1
2
343
This new version of Alex Wang is hilarious!!
This quoted post is unavailable.
1
173
This, not search, is the real Google killer. Truly helpful sidekick that pushes the right information at the right moment!
Now in preview: ChatGPT Pulse This is a new experience where ChatGPT can proactively deliver personalized daily updates from your chats, feedback, and connected apps like your calendar. Rolling out to Pro users on mobile today.
4
5
446
Not @Benioff with the YC founder always be launching vibe!!
Welcome Slack ITSM by Salesforce.
4
457
Hot take: The "founding creator" trend is a trap most startups are walking into blindly. Every startup now ships cinematic launch videos because when software takes weeks to build, distribution buys you some moat. But we're confusing skills with DNA. 🧵
2
4
324
Your "founding creator" will quit the moment they hit 100K followers. They're building THEIR brand, not yours. The content that goes viral about your startup? It's actually about them.
1
2
157
They can absolutely be founding team members. Just be honest about the trade-off. Give them equity, pay them well, but don't expect them to stick around during the boring months when there's nothing sexy to post about.
3
126
Conferences can be pretty effective channels to land and close customers. Avoid at your own risk!
Huge founder time-wasters – Conferences (unless you’re speaking) – constant networking – Meeting VCs when you’re not raising – Entertaining M&A talks when you’re not selling You’re paid to grow your company, not to network with people who are paid to talk.
4
568
every time you email the Rippling team, they hand you off to your new account manager quite frustrating tbh
1
6
684
startup hiring has always been go slow to then move fast. for the first time now - if you just need execution bandwidth — there are real alternatives
1
8
554
A picture is quite literally worth a 1000 words
images in chatgpt are wayyyy more popular than we expected (and we had pretty high expectations). rollout to our free tier is unfortunately going to be delayed for awhile.
2
746
This 👇
georgia tech might be america's u of waterloo
1
410