allie lawsen @lexlawsen.bsky.social · 24/04/2025I could have titled my new blogpost "The one skill that makes or breaks your career." I didn't, because that really isn't my style. But I do think the post contains a lot of the best general career advice I can give. 🧵 below lawsen.substack.com/p/four-and-a...lawsen.substack.comFour (and a half) Frames for Thinking About OwnershipIt's hard to write a guide to being good at ownership. 100
allie lawsen @lexlawsen.bsky.social · 24/04/2025Interesting... Proxy-generator is here in case you're still using the other place: substack-proxy.glitch.me 000
allie lawsen @lexlawsen.bsky.social · 09/04/2025If only this was the AI that had been asked for tariff setting advice. 110
allie lawsen @lexlawsen.bsky.social · 03/04/2025What deep research queries have you tried and been disappointed by? Thinking about writing a post on how to use it well and it would be good to have interesting examples. Reply with the prompt and model you used. 042
allie lawsen @lexlawsen.bsky.social · 01/04/2025What's the difference between good and *great* prompt engineering? 🧵 ↓ 100
allie lawsen @lexlawsen.bsky.social · 24/03/2025What does it take for people to actually, reliably use a whistleblowing function? 1/8 120
allie lawsen @lexlawsen.bsky.social · 06/03/2025Just published a new post on the Claude Projects I'm currently maintaining. I've had a lot of interest in how I'm using AI, so wrote this breakdown. 110
allie lawsen @lexlawsen.bsky.social · 26/02/20251. I'm unironically excited about Anthropic's Pokémon eval. It's a step towards the kind of thing I was hoping we'd get in response to this section of our recent RFP. 151
allie lawsen @lexlawsen.bsky.social · 26/02/20251. I wrote this a while ago, but it's come up in a few conversations since so it seemed worth a short thread. 100
allie lawsen @lexlawsen.bsky.social · 11/02/2025New post open.substack.com/pub/lawsen/p...open.substack.comMy current writing workflow.As of Feb 2025 110
allie lawsen @lexlawsen.bsky.social · 06/02/20251/8 Really happy to share Open Phil’s new Request For Proposals to improve AI capability evaluations. It's been a big project for me and my colleague Catherine. Getting evaluations right is crucial — and we're ready to fund serious work to make that happen. 🧵 161
allie lawsen @lexlawsen.bsky.social · 01/02/2025Got a fully adjustable, ergonomic, steel case chair for working from home and it sucks. My back still hurts after 5 hours working hunched over my laptop on the sofa because I didn't want to go upstairs. 020
allie lawsen @lexlawsen.bsky.social · 28/01/2025TFW you realise anti-sycophancy training still has a way to go because your conspiracy theorist relative texts you that Claude agrees with them "once they supply it with the facts the MSM is missing" 030
allie lawsen @lexlawsen.bsky.social · 24/01/2025OpenAI's operator, from the sound of it, barely works when it comes to bunch of things. Luckily, as we all know, it's really hard to go from 'barely works' to 'works' to 'superhuman' in AI, especially once you have the basic set up that gets you to 'barely works'. 090
allie lawsen @lexlawsen.bsky.social · 17/01/2025Intelligence has to be from the Jacques region of France! What you're talking about is just sparkling matrix multiplications! 150
Reposted by allie lawsenAndy Masley @andymasley.bsky.social · 13/01/2025Wrote a long rambling post about why individual AI use isn't bad for the environment substack.com/home/post/p-...substack.comIndividual AI use is not bad for the environmentAnd a plea to think seriously about climate change without getting distracted 0196
allie lawsen @lexlawsen.bsky.social · 07/01/2025Training a flexible, general-purpose reasoner that can succeed despite unexpected obstacles seems pretty hard. Worryingly, training a flexible, general purpose reasoner that can succeed despite unexpected obstacles *except when those obstacles are humans trying to stop it succeeding* seems harder. 091
allie lawsen @lexlawsen.bsky.social · 07/01/2025New blog post open.substack.com/pub/lawsen/p...open.substack.comSay It Twice, Write It DownFollowing my own advice 030
allie lawsen @lexlawsen.bsky.social · 30/12/2024Would you be more likely to read an essay entitled: 1. "Just" has no explanatory power 2. "Just" is a semantic stop sign 3. Un-Just dismissal ? 110
allie lawsen @lexlawsen.bsky.social · 30/12/2024Have you ever actually seen a novel consideration from a reply which starts "have you considered..."? 110
allie lawsen @lexlawsen.bsky.social · 30/12/2024Yeah but it's obviously going to slow down now. Even in 10 years I doubt even the best model will get another 5% of improvement on the left hand test. Would give 1000-1 odds against it. 040
allie lawsen @lexlawsen.bsky.social · 19/12/2024bsky.app/profile/lex... Footnote to this, about a different set of reactionsbsky.appallie lawsen (@lexlawsen.bsky.social)Quick reactions to some of the other reactions I've been seeing to Anthropic's recent paper on "Alignment faking in LLMs" 1/9 120
allie lawsen @lexlawsen.bsky.social · 19/12/2024Quick reactions to some of the other reactions I've been seeing to Anthropic's recent paper on "Alignment faking in LLMs" 1/9 110
allie lawsen @lexlawsen.bsky.social · 17/12/2024His palms are sweaty, hands weak, fingers heavy they tremble as he types out the prompt Will's spaghetti He's nervous, but on the surface he looks calm and ready to type out "Deep learning is hitting a wall" x.com/deepfates/st...x.comx.com 020
allie lawsen @lexlawsen.bsky.social · 14/12/2024Appreciated this from @thezvi.bsky.social Full post here: www.lesswrong.com/posts/HfigEy... 021
allie lawsen @lexlawsen.bsky.social · 11/12/2024I'm worried about takeoff speeds, said Tom, swiftly 140
allie lawsen @lexlawsen.bsky.social · 11/12/2024It's 2028. Takeshi's Castle now has robot participants. They aren't doing as well as the best humans. Looks like deep learning is hitting a wall. 130
allie lawsen @lexlawsen.bsky.social · 11/12/2024Gemini's live mode is crazy good. It handled me interrupting it, let me pause to think, and the voice quality is incredible 110
allie lawsen @lexlawsen.bsky.social · 11/12/2024There were many awesome things about working at @80000hours.bsky.social, predictably including seeing people I'd advised get involved in cool projects, but it's striking to me how much joy I'm still getting from this nearly 18 months after leaving. My advisees just keep killing it! 1100
allie lawsen @lexlawsen.bsky.social · 10/12/2024OK, who the hell is Jacob and what does his Pie shop have to do with Pot Limit Omaha? 220
allie lawsen @lexlawsen.bsky.social · 10/12/2024Addendum to this thread: Also, don't test an early checkpoint that's way dumber and consider that "good enough". 000
allie lawsen @lexlawsen.bsky.social · 09/12/2024Writing this was a huge win for the "if you find yourself saying something a lot, write it down" principle. 050
Reposted by allie lawsenAlexander Berger @albrgr.bsky.social · 07/12/2024But the people I’ve met who are most acutely, painfully paying attention to the fact that you can save kids’ lives for ~$5K, and what that means about the world’s inequality and gross disregard for the poor, tend to be GiveWell donors. 1374
allie lawsen @lexlawsen.bsky.social · 07/12/2024It's like they made a list of every single safety-related commitment or effort they'd ever engaged with, and are now working through it trying to remove every one. www.ft.com/content/2c14...ft.comOpenAI seeks to unlock investment by ditching ‘AGI’ clause with MicrosoftStart-up discusses removing provision to protect powerful technology from being misused for commercial purposes 3113
allie lawsen @lexlawsen.bsky.social · 07/12/2024"When I tweet out engagement bait, I put it in quotes to seem more insightful." 060
allie lawsen @lexlawsen.bsky.social · 05/12/2024With all the time he's spending off Twitter, can you believe how good Julian has got at singing? www.instagram.com/reel/DC-ybqN...instagram.comLogin • InstagramWelcome back to Instagram. Sign in to check out what your friends, family & interests have been capturing & sharing around the world. 010
allie lawsen @lexlawsen.bsky.social · 04/12/2024What podcasts should go into a (forthcoming) EA/AI podcast tier list? 340
allie lawsen @lexlawsen.bsky.social · 04/12/2024Enjoyed this from @garrisonlovely.bsky.social open.substack.com/pub/garrison...open.substack.comIs the AI Doomsday Narrative the Product of a Big Tech Conspiracy?A close reading of what tech titans actually say on the matter 020
allie lawsen @lexlawsen.bsky.social · 04/12/2024Biosecurity comms: █████████ (QT @sebk.bsky.social on the other app.) 0130
