Andy Vandervell (@andyvan) header
Andy Vandervell (@andyvan) profile picture

Content for @posthog. Recovering journalist. Optimist.

London
Joined January 2009
Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)Photo by Andy Vandervell (@andyvan)
Andy Vandervell retweeted
Dev chores stink. Automate them.
4
4
65
3,309
Andy Vandervell retweeted
Self-driving is out of beta - we got pricing wrong at the start. we were charging $15 per pr, and reports were free - agents do most of their work while creating reports. that's where the value was, and what people actually cared about - now every run is billed, including reports, based on how long the agent spends in the sandbox and how much context is read
Hog, take the wheel. Self-driving is now GA. Scouts read your events and errors, write Reports summarizing them, and then agents turn them into PRs.
15
10
5
88
21,399
Andy Vandervell retweeted
Gamers, rise up. We're building an Unreal Engine SDK. Not ready for prod (yet), but already supports - Product Analytics - Session Replay (on Mac) - Feature Flags
9
3
4
130
9,903
Andy Vandervell retweeted
Confirmed: brand.posthog.com/hoggies works on Wii browsers
8
4
4
63
14,418
Andy Vandervell retweeted
The cost of coordination is often way larger than the cost of some duplicative work across an org. This came up last week in our slack when someone cross checked our teams' Q4 goals. - it costs a lot in overhead to silo tasks - in most cases different teams' work will be better if it bleeds over anyways - it's dumb to spend too much time directing people
1
9
828
Andy Vandervell retweeted
oh my god are you guys reading this
6
1
16
1,713
Andy Vandervell retweeted
If you want to understand why it’s so hard to pause the AI race, the best place to start is Prozac My latest for @skynews:
1
26
10
108
26,070
Andy Vandervell retweeted
Having all your product and business context in one place makes it easier for you and your agents to do analysis and make decisions. Our context warehouse makes this easy. To show you how, here's 11 sources we sync into our context warehouse at PostHog and what they let us do.
8
1
2
51
7,031
Andy Vandervell retweeted
Get the answers you need from PostHog in Slack, and set your status to "🚌 Commuting" until EOD
2
2
1
46
6,200
Andy Vandervell retweeted
Haven't had time to look at this in detail, but just on the face of it - 6.6m litres a day is 0.25% of England's non-household potable water supply, a little bit more than the amount consumed by the postal and courier sector Add in households and data centres account for around 0.05% of the total public water supplied By comparison, leaks account for 19% Somewhat confusing why data centres that use 0.05% are unforgivable but leaks that consume 400 times that amount are forgiveable - but I guess that's what you expect from the trade body for water companies 🤷
Replying to @neill_pippa
Datacentres in England use an estimated 6.6 million litres of drinking water every day. Jon Chappel, Water UK’s deputy director, said this was “unforgivable” and a “genuine failing of government”. theguardian.com/environment/…
Readers added context they thought people might want to know
Water lost to leaks in England and Wales was around 3 billion litres per day in 2024-25, according to government statistics. Data centres currently use less than a quarter of one percent of this amount. gov.uk/government/pub…
18
66
2
508
40,792
Andy Vandervell retweeted
Sky's @rowlsmanthorpe looks at what the appointment of Kanishka Narayan as AI minister could mean, and why he thinks this feels like 'a demotion' For more tech news: news.sky.com/science-climate…
19
19
4
159
189,449
Andy Vandervell retweeted
How @posthog went from this to this 🧵 1/7
41
10
12
558
73,793
Andy Vandervell retweeted
New hog merch alert
3
1
5
26
6,649
Andy Vandervell retweeted
11 thoughts on the AI situation now Fable 5 and Mythos 5 have (sort of) been released 1. There is now a de facto licensing system for AI in the US and by extension the rest of the world 2. The labs are doing everything they can to make the licensing framework as formal as possible - that's what Anthropic's calls for "common standards" and "frameworks" are all about 3. While this might not seem like the language of the Trump administration, it has a huge incentive to find some kind of framework because the financial implications of getting it wrong are so enormous 4. And I mean enormous. The US stock market is propped up by AI yet the US government is delaying/nerfing the best models, making it very hard for them to compete against Chinese labs. Surely this cannot last 5. So you need to improve licensing. But any kind of licensing regime will inevitably slow down model releases and possibly make them less powerful 6. How to square this circle and stop licensing damaging the US labs? The conclusion is obvious - ban or heavily regulate Chinese AI or even open source AI 7. Anthropic has clearly clocked this because Dario has been lobbying for open source regulation 8. Of course this would be v difficult to enforce but it's not impossible as some people make out. There's a bill in Congress calling for a ban on Chinese AI in the govt supply chain - pass that and people will fall in line 9. I expect this struggle to go on for some time and be very chaotic but there are models for where this might go - most obviously medication 10. Good version: AI is centrally regulated but open source is allowed under certain conditions, leading to the kind of split you have with generics vs big pharma 11. Bad version: open source is totally banned and the labs can charge whatever price they want - "illicit" AI is therefore highly desirable but is constantly being cracked down on in a heavy-handed way, like Napster or torrenting before legal streaming 12. Given the undesirability of that last scenario, there is a big opportunity for non-Chinese open source AI right now - perhaps an opportunity for EU/UK/Canada etc? 13. But mostly for the rest of the world, this is very difficult. Right now the licensing system looks like a lot like what Sam Altman calls "the government picking the customers" - and those customers are all American 14. There is an urgent need for a global settlement involving both China and the US or this could get extremely chaotic and that would not be good. The labs have already said they'd support this. Who can lead it? Just thoughts and not a statement of my preferred outcome. Interested in hearing where I'm wrong – although of course if you think I'm right please share this with fulsome praise
10
4
2
27
7,738
Andy Vandervell retweeted
I'm happy to announce that after completing a thorough investigation, I have decided to join @posthog. See results bellow 👇 youtu.be/guzM7G1HDIs
17
6
60
12,592
Andy Vandervell retweeted
Hear this sentiment a lot
I cannot describe the fury that I will have if any UK PM changes leads to a reshuffle in the ministers in Dept Science & Tech (DSIT) They have been doing a phenomenal job - specifically I’ve never met anyone with a bad word to say about @KanishkaNarayan Consistency in leadership for tech policy is essential over the next few years.
1
1
21
7,771
Andy Vandervell retweeted
In 2025, we interviewed 2,431 candidates and we observed two things: 1. Over 70% of the people we hired just applied and had no referral. 2. Candidates who asked the best questions were more likely to get hired. So we made this video about the questions we wish people asked more. Full vid on YouTube: youtube.com/watch?v=KVMZgY2K…
5
4
2
142
30,256