Neil Vachharajani

Congratulations! You're now a middle manager. Every IC, in every function (sales, marketing, ops, finance, HR) just got promoted. HR didn't send a note. Your comp didn't change. But here we are. Your direct reports: a fleet of AI agents. Here's the trap: they will never quit because you micromanage them. No passive resistance. No "exploring other opportunities." Just quiet compliance forever. Micromanagement has always been a tax on output. The feedback loop was humans pushing back. Agents broke that loop. The people who win won't be the ones doing the most. They'll be the ones who manage at the right altitude: delegating clearly, defining outcomes not steps, trusting their agents to execute. New title. New skills. New tools. We're building the infrastructure to help people actually run agent fleets, not just babysit them. Curious? Reach out at run.so

Agent memory is still an unsolved problem. Most work on memory falls into one of three buckets: 1. RAG against a static corpus 2. Agentic RAG - a tool call to retrieve from dynamic memory 3. Better indexing for either of the above But none of these are how human memory actually works. Human memory isn't a slow-moving corpus. It's fast-changing. And critically, we don't *fetch* memories by posing carefully curated queries. Subconscious association surfaces the right context at the right time, automatically. Imagine a coworker who acknowledges something you said, then forgets it 30 seconds later. It'd feel broken. Yet that's exactly how most agents behave today. The hard problems few people are talking about: fast-changing memory that stays coherent, and associative retrieval that doesn't require the agent to "know what to look for." This is what we're working on at Run. If these problems interest you, we're hiring.

@mvachhar My guess is because the installer inspects your system to figure out what to download. Users don't have to answer questions like whether they use Intel or Mac silicon, etc.

As I set up my new Mac, I do not understand why so much software makes me download an installer that then downloads the actual software to install. Why not bundle the two together?

@ravivenk I do think it's a problem and I'm actively investigating the space! Let's talk more!

All these auto code generating tools and companies are great, but but hear me out. I rather want a tool that helps me debug and fix code when shit hits the fan and production is down. Way too many logs to look and correlate. Freaking nightmare in every company I ever worked.

@johnfreehayes Imagine if you had to pay per Google doc. I wish there were a silver bullet that computed how much value users are getting...

@johnfreehayes Seat based licensing is terrible and encourages "cheating" and non-optimal use of tools, but so does usage based pricing, especially when there are high and low value uses of a product. IMO you don't want your users stressing about whether this is a high or low value use.

Everything will move to usage based pricing: 1. Seat engineering is the most tedious part of negotiation - only the total matters but making two levers complicates it. 2. B2B should be motivated to drive use like B2C. 3. You want neutrality between human and robot users.

@mvachhar @googlecloud #chatgpt is a amazing. I tried it to summarize question and answer threads and it worked remarkably well.

#ChatGPT just saved me a ton of time. I asked it to write terraform to create a Let's Encrypt cert and add to @googlecloud certificate manager. It's code was full of errors, but it the providers and wiring were right. A quick look at the docs, some fixes, and I was done.

@DBArgenis 3/ I'm not sure we'll ever eliminate those differences or even want to.

@DBArgenis 2/ And few talk about the innate differences that exist because the families of children are different and those differences end up being important and generational (not just the wealth and income of those families, but the values more importantly).

@DBArgenis 1/ It is a good thread, but I'm not sure I agree with the two camps being the only two camps. I think the fairness folks can be divided into equality of outcome and equality of opportunity folks.

@ravivenk Feel free to drop some off 🙂

My son is getting good at this, he made this for breakfast today. Tastes yummy! 👍🏽👍🏽

@DBArgenis @SQLSteinar @elonmusk Oh, and the Tesla repair center said that all of their replacement wind shields were defective :-/

@DBArgenis @SQLSteinar @elonmusk Safelite told me it'd take 5 weeks to get the windshield. Which model do you have? I'm trying to get it for a new Model S

@richhankins I feel like this is an @rolanddreier tweet 😛

@rolanddreier @PureStorage CloudArray and CloudBladr?

Amazing — “all-flash” went from revolutionary to not even worth mentioning — @PureStorage Q4 earnings PR doesn't say “flash” at all, aside from tiny handful of references to existing FlashArray/FlashBlade products (& I bet they wish those could rebrand) prnewswire.com/news-releases/…

@tarasglek I'm hoping any friends and family are safe!

Russians are shelling Mariupol, casualties among civilians and military.

@DBArgenis This is horrifying. Keeping my fingers crossed I don't need any medical care between now and Feb 1.

@nvachhar nbcbayarea.com/news/coronavir…

The covid transition is here.

@DBArgenis Yeah, I can't find that on DPH's website. Seems crazy if it is true.

@nvachhar Of the news?