Prak, Inc

Insane stat of the day: California almonds use roughly 3–5.5 million acre-feet of water per year, depending on methodology. That's ~4-7x more water than all data centers in North America used combined in 2025.

your pricey upcharge for real functionality in your GTM saas product is no match for a base tier plus Claude in Chrome.

Everyone talks about MCP like it’s magic infrastructure most MCP implementations today are actually a disaster for enterprise scale. teams are shoving hundreds of API endpoints into giant MCP servers and blowing up context windows + token costs. one company told me their platform has 500+ endpoints. raw API-to-MCP conversion would completely break the agent experience. the future is not: “here are 500 actions an LLM can take” the future is: “here’s the exact outcome you wanted in one call” that’s a very different product design mindset than SaaS APIs were built for.

Most companies don't kill ideas by saying No. They kill them by making you ask 14 people for permission. @dickc fixed this at Twitter when he was CEO with one rule: only your direct manager can block you. "Experiments started flying out." That's bias to yes.

Make sure your AI tools are two-way- reducing friction for both advisors and complaince- #OperativeAi for #RIAs praktikantai.com/meeting-next-s…

A critical question in agent design is “how do we build agentic workflows so humans are given significant, interesting, or variance-producing decisions as they come up in the work?” A Claude-run company has no source of competitive advantage compared to other Claude-run firms.

Your best employees will still be your best employees in the age of AI, but they won't be pulling all nighters. #OperativeAI for #RIAs praktikantai.com/heres-how-to-s…

For those not on Enterprise (or trying not burn token on a whole new draft)

At a recent Anthropic user conference they noted that for every dollar spent on their software, customer spent $8-$12 on services. You’re not bullish enough on AI services.

Addepar is experienceing a longer than normal data delay with the custodian feed from Schwab. No affected month end reports will be run until rectified. your Prak jobs will remain in queue. #OperativeAI for #RIAs

cc @creatorsports_

My $AAPL watch is really getting good. #LetsGoFlyers #StanleyCupPlayoffs

@hwchase17 Deep agent validating track record presented in data room by PE manager- by finding third parties associated with the deal that can quickly confirm high level qualitative and quantitative details.

I’m going to build a bunch of deepagents examples (using deepagent deploy) over the next few days What examples would people want to see?

Dear $MSFT- it should only be one click to switch Teams environments, not two. pls fix.

Ok we're shipping something this weekend that we've REALLY needed 💯An Agentic API Grader. An objective view of APIs from an agent perspective. For builders of B2B apps especially. Which APIs are easiest to use for AI Agents? The hardest? We've got 100 APIs already loaded in, especially ones forces on B2B apps, and will keep working on it: @stripe: A (95) @Adyen: A- (83) @RevenueCat: A- (82) @linear: A- (80) @clay: B (73) @brexHQ: B (72) @HubSpot : B (70) @tryramp: B (67) @Gong_io: B (60) @ZoomInfo: C+ (58) @NotionHQ : C (52) @Marketo: C (50) @Workday: D (38) And we've got full details on why. We're entering the age where you should pick your vendor based not on what your humans need. But what your agents need. For Team SaaStr AI, we're already there. Much more at link in 🧵

Doing for nuclear what Henry Ford did for automobiles. @AaloAtomics is going to be a generational company.

Stop asking your best models to explain the plot of The Boys Episode 4.