richardw

22 ms·

1. 271.▲
   by richardw 2y ago

   I’m not sure how it’s going to be any different for us. We keep saying we’ll be using these tools, but not understanding. The tools aren’t just tools. When they’re smarter than you, you don’t use them. The more you try to enforce control, t
2. 272.▲
   by richardw 2y ago

   According to the repo README: 'java', 'markdown', 'python', 'php', 'javascript', 'c++', 'c#', 'c', 'typescript', 'html', 'go', &
3. 273.▲
   by richardw 2y ago

   I’d be interested to know if that trade off ends up better. There’s probably a lot of useful training that transfers well between languages, so I wouldn’t be that surprised if the extra tokens helped across all languages. I would guess a to
4. 274.▲
   by richardw 2y ago

   I think you have a lot of potential customers who know they have a multi tenant challenge but don’t know that they have a “don’t roll your own” challenge. Most multi tenant systems fail open rather than fail closed and leak data very easily
5. 275.▲
   by richardw 2y ago

   I think the term is simply ownership. You choose what to own, and often good founders (and honestly managers) own anything that is important. There’s no “delegate and go home”, there’s “trust, but verify” and when verification fails, you ge
6. 276.▲
   by richardw 2y ago

   Checks out from this old account of a BillG review: https://www.joelonsoftware.com/2006/06/16/my-first-billg-rev...
7. 277.▲
   by richardw 2y ago

   Yeah but when your company gets a real challenge, you find out if you’re filled with builders or comp optimisers.
8. 278.▲
   by richardw 2y ago

   Recommend just letting people have their one day in the sun. We’ve become less the site of builders as the red team for testing your launch.
9. 279.▲
   by richardw 2y ago

   Well, adjusting nudges in aggregate but diced in various ways. Measured very much not in aggregate. We’d see positive and negative outcomes roll in over multiple years and want it per identifier (an individual). I’ve heard of companies gene
10. 280.▲
    by richardw 2y ago

    I designed an internal system that optimises for long term outcomes. We do nothing based on whether you click “upgrade”. We look at the net change over time, including impact to engagement and calls to support months later and whether you l
11. 281.▲
    by richardw 2y ago

    Ah, text. It's so shit for conversation. I feel like I'm speaking to a smart person and we'd do a lot better over a beer, and we're probably a lot closer in opinion somewhere. Maybe. > Surely ChatGPT 5 will prove one
12. 282.▲
    by richardw 2y ago

    > AI is no more familiar with our lives on Earth than it is Frodo's life in the Shire. It can only render solutions amenable to the perverse portrayal of reality that text provides for it. What’s the relevance? Let’s be open to the
13. 283.▲
    by richardw 2y ago

    I wish people would stop projecting decades into the future using advanced autocomplete as the definition of AI. We were way fucking surprised at the output of token prediction but somehow assume the surprises are all gone and we can predic
14. 284.▲
    by richardw 2y ago

    Quite a few responses to your questions based on current state of AI. Firstly, this is never going backwards. It will only get more capable. There will likely be algo changes that unlock new capabilities over time. There are definitely area
15. 285.▲
    by richardw 2y ago

    This is still backed by Mike Cannon-Brookes from Atlassian, isn’t it? Not mentioned in article but: “May 2023, a consortium led by Cannon-Brookes' Grok Ventures won the bid to acquire Sun Cable,[7] with the takeover finalised on 7 Sept
16. 286.▲
    by richardw 2y ago

    I very much feel you. Choosing customers is crucial. Patio11 has some great write ups on the type of customer who will suck your life force: Retail, not much money, not very busy. Rather have business clients who are less price sensitive an
17. 287.▲
    by richardw 2y ago

    I stop it from coding, just ask it to brainstorm with me on a design or plan. After a few iterations it knows what I want and then I ask it for specific outputs. Or I code those myself, or some combination. I’ve found that it’s sometimes am
18. 288.▲
    by richardw 2y ago

    Humans also remember and update their model in real-time (with various caveats obviously). LLM’s forget when you close the window and need 1000 examples before they learn anything.
19. 289.▲
    by richardw 2y ago

    I’m spending too much time at home. Getting a lot done but serendipitous meetings and network building is definitely impacted. Time with my daughter is way up, before and after school.
20. 290.▲
    by richardw 2y ago

    I’m not sure anything else was material given that the machines were bricked and client roll-out approaches were evaded by Crowdstrike. What client actions would have helped? Surely someone is looking at a class action? People died. The con
21. 291.▲
    by richardw 2y ago

    I was part of a team for 12 years and watched it for longer, that helped a company on a journey exactly like this. I don't really want to get caught up in phone calls because I need to focus, but here's some of it. Rough version:
22. 292.▲
    by richardw 2y ago

    It's not a unique model - that's what e.g. Databricks does too. You pay for the resources, which means you can deploy them with whatever security model you require (good for enterprise, with all the firewalls and secrets and secur
23. 293.▲
    by richardw 2y ago

    I had contacts when I was younger. I never figured out how to not have them hurt when putting them in the day after working with chili. Gloves are a good idea but this was usually some random place I’d be cutting them with friends.
24. 294.▲
    by richardw 2y ago

    Sure. My base case is that LLM’s find uses and that compute gets cheaper and more efficient, which helps make more business cases possible. Investing in current state GPU’s doesn’t seem that useful in that world. Nvidia is priced for a mono
25. 295.▲
    by richardw 2y ago

    Or they invested in GPU’s to mine bitcoin and the ASIC’s are just starting to arrive. At least for inference, GPU’s are not the end game. And who knows if training will continue to need them. If so there could be another round of investment
26. 296.▲
    by richardw 2y ago

    Here’s another: A16z building a stash of GPU’s: https://www.theinformation.com/articles/andreessen-horowitz-...
27. 297.▲
    by richardw 2y ago

    Dump the concrete! Put baby powder around the scene with bio hazard warnings!
28. 298.▲
    by richardw 2y ago

    They’re branding experts. Getting published by Nature is far more valuable to someone compared to (all the other options). Humans prize exclusivity. We’ll hand our time and resources to people who can give us a tiny bit of it. The harder we
29. 299.▲
    by richardw 2y ago

    Despise Godaddy. And google domains sold to Squarespace, who makes a transfer take up to 15 days. It’s faster to carve the document from wood and send it across the world than it is for them to automatically click the button to let go of a
30. 300.▲
    by richardw 2y ago

    I think a good place to start is “anyone in the world who wants to learn and is hindered by their school and teachers”. Not “this must work in every American school from the start”. A smart kid with a phone in Kenya should be able to have a