Built a fun side project today. WhatWouldClaudeUse.com shows which vendor/tech LLMs recommend for vibe-coding use cases. Some categories are competitive (see: email APIs, below) and others are clean sweeps (@sentry in observability, @vercel in hosting, etc.).

Neon acquisition looks smart for Databricks- the managed database of choice for the world's smartest model! WhatWouldClaudeUse.com

Email my AI agent at: analyst@autonomousresearchcorp.com Any requests. Anything you'd ask a human analyst to do. Attach stuff to the email. Paste images. Rant into the text box. Be as sparse or as detailed as you'd like. 5 free requests. No strings attached. I can't guarantee satisfaction. But hey - it's free. What do you have to lose? Share the email to anyone in your life that would love to test this out as well.

One-click domain name setup between @Railway and @cloudflare domains is 👨‍🍳💋

Starting from base rates is the best way to predict the future… and the base rate for bursts of efficiency causing unemployment is low (maybe zero). Ironically a big risk factor is regulation/government intervention preventing labor adaptability. See: electrician shortage

It would be crazy to promote an employee based on how smart they are. Yet most companies using AI are paying for intelligence, not for results. Companies need to evaluate agents as they do human employees - based on impact.

Ramp is continuing to just kill it on product releases.

If you’re the kind of person who thinks the world is somehow worse off because these people did this, I don’t know what to tell you. What a brilliant display of humanity. Makes me so happy

This is sick- and increasingly critical to have/deploy.

It’s amazing to me how many investors think the current run on chips/memory/etc. is driven by enduser demand. $100b of OAI/Anthropic ARR = ~$30b of cost which is ~$150b of capex. Drop in the bucket. Inference for post-training/RL seems to be the real hog today.

I have a super-talented friend who was early at one of the hottest startups in NY and is now looking for a cracked technical co-founder. Couldn't be a better opportunity- the idea is great, she's great, the timing is great. If you or someone you know is truly special, DM me.

Entry-level headcount grows even faster, 12% over two years. This is our first evidence that high-AI adopting firms are hiring different kinds of employees. We believe they are selecting for a new set of skills, specifically, people who know how to use AI and use it well. Entry-level workers, especially recent graduates and college students, are a natural place to look.

I'm increasingly convinced that for many applications, per-user pricing will still predominate in five years. In some cases, we might snap back from value-based/consumption-based pricing to per-user pricing. 1) Many use cases simply won't require that much intelligence, so existing SaaS pricing will largely hold. 2) Per-user pricing solves the inference moral hazard issue. A vendor charging you a flat price per user can be trusted to route queries to different models on your behalf, but a vendor that is in the "token path" is incentivized to drive higher spend. It's a law of physics. 3) In many cases, the right proxy for "how much are we willing to spend on AI in this area" will map to "how much are we willing to spend on people in this area." So far, the evidence is pretty strong that AI and people are complements, not substitutes (obviously there are exceptions). So long as they are complements, AI spend as a proportion of human spend is relevant to pricing/value- and # of humans is the best/simplest proxy we have for human spend. No doubt there will continue to be lots of pricing model experimentation, but the blind "all SaaS will become consumption based/value based" takes are probably wrong.

It's a common myth that a pitch is about delivering information. In reality, the information is almost always available somewhere- online, etc. Pitches are about capturing attention with a combination of presence and storytelling. It's the same skill as teaching, actually.

Devin Fusion: - solves cost - extremely smart - a new era of multi-model agents It’s very fun to use. Check it out!

Certain formats are good (TV) to great (feeds/OG search) for ad insertion. Others are not- in part because the context is tough (ads next to deeply personal convos, etc.) and in part because you aren't "selecting" from a menu of options the ad can sit next to.

Life hack: Get all of the important events for people you love (wedding anniversaries, birthdays, etc.) into a color-coded Google sheet. Then check it once a year and plan around special events. Even better yet, you can share it with the rest of the fam!

Your company doesn’t have to 996 to win, but if you message your boss Saturday morning and they don’t reply until Monday morning, yall are ngmi.