Beyond the tech world? There is nothing. 🌐

Joined June 2023
Photos and videos
Feels right tbh. Open models are moving properly now. Once they’re good enough, I’m not paying top-end API prices unless I have to.
Kimi K2.7 has just been released & it's a really good model! Open models from China are catching up very fast! By the end of the year or at the latest, early next year, we’ll likely have a Fable 5-level open-source model available at a much lower cost! AI advances cant be stopped
1
AIKeaton UK retweeted
AI coding founders like @matanSF are always my favorite people to listen to. Epic entrepreneurs. I think it’s because (1) they sell highly technical products to technical buyers. So their day job is talking about complex engineering non-stop, and the clarity of thought they need for this filters into how they talk about everything. And because (2) AI coding is so fast-moving and competitive, the founders of AI coding companies evolve into apex performers. Factory, Cursor, Cognition, plus OpenAI, Anthropic — in terms of talent, fundraising, revenue growth, each of these is one of the best businesses in human history. (@generaltxn is proud to serve a few of them as our customers!) The people leading these companies are like athletes who train at high altitudes. The environment forces them to operate at a much more elite level even than a typical successful founder. I admire them all so much.
Watch @HarryStebbings ask @matanSF all the difficult questions about commoditization across the AI stack, Factory's controversial culture, or whether US startups should use Chinese models. The 20VC episode is out now!
1
18
2,621
AIKeaton UK retweeted
Alexandr Wang offers a contrarian take on management: the more brilliant your people, the less you should tell them what to do. Most leadership instinct says that as you hire better people and take on bigger problems, you direct more closely. Wang's approach to running MSL goes the other way. "In general, in terms of my management philosophy for MSL, [it] is not to boss people around," he says. @alexandr_wang points to a line from Steve Jobs that flips the usual logic: "Most companies hire people and tell them what to do, but we hire people for them to tell us what to do." That inversion sits at the foundation of everything. For Wang, the entire bet behind MSL and TBD rests on it: "That is like pretty core to the entire thesis of TBD and MSL and how we've built it… we're going to hire brilliant researchers and create the best environment for them to do the work of their careers and the work of their lives." So instead of positioning himself as the person with the answers, he positions himself as the person who removes everything standing between great researchers and great work: "Long story short, I'm not trying to boss anyone around. Actually, I'm trying to create the best environment for researchers to do incredible work."
1
20
51
3,856
AIKeaton UK retweeted
Harrow School Dubai Set to Open This September as First Middle East Campus Harrow School will open its first Middle East campus in Dubai this September, bringing its 450-year British education legacy to the UAE. More updates at #TheExpatStory #DubaiEducation #HarrowSchoolDubai #UAESchools #DubaiNews
1
109
My take: you can't build a global AI business on access that gets revoked overnight. All the blame-game stuff is secondary to that.
Blame Anthropic? Blame Howard Lutnick? Blame Andy Jassy? Blame Canada? First emergency @aidailybrief episode since the sama ouster and let me just say... There's plenty of blame to go around.
2
Bad day for vibecoders, but I think this gives open models a bigger market. Gate the top model and everyone else just pushes harder to catch up.
America is hoarding Claude I think this will give room for allot of other llms opencodes to step in indirectly creating a bigger market and because a lot of llms won’t still match up to Anthropic mythos Mythos takes the not everyone can have access and expensive Others are going to keep trying to meet up like Apple and the rest of the android industry
1
AIKeaton UK retweeted
Jun 13
Boris Cherny - the father of Claude Code - just explained why one developer with AI now does the work of a team of 10 and keeps all the money here are 4 moments from his talk that will change how you think about code: 00:00 - Anthropic has no more manually written code anywhere at the company. all SQL is written by models. everything is built by models 06:04 - every night he has a few thousand agents doing deeper work while he sleeps. in the morning he just checks the results 08:19 - last week there was a day when he closed 150 pull requests in a single day save this video - the people who figure this out first will be earning 10x more than those who are still waiting for the right moment
18
9
91
11,307
AIKeaton UK retweeted
FRS is not just an inference trick. The discovered latent noises can be distilled with Diffusion Steering via Behavioral Cloning (DSBC), training lightweight steering policies in under a minute, or used to bootstrap RL with semantic guidance. On real DROID tasks, DSBC significantly outperforms standard behavioral cloning, while FRS-bootstrapped RL succeeds on tasks where conventional RL makes little progress.
1
1
4
253
AIKeaton UK retweeted
jeff dean just told uw grads that "ai is an incubator, not a substitute" as his top career advice for 2026. headlines: google's chief scientist endorses ai collaboration. it isn't. dean just validated harness theory from the most credible voice in production ai. not replacement. amplification. ai doesn't eat your job. it becomes your cognitive exoskeleton. businessengineer.ai/p/has-mi…
Jeff Dean's Three Rules for the AI Era: Why Google's Chief Scientist Just Endorsed Harness Theory fourweekmba.com/jeff-dean-th… #AI #Google #HarnessTheory
71
AIKeaton UK retweeted
If your practice does not expose weaknesses specific enough to correct, you are just practicing staying the same.
1
25
208
3,005
Not even that surprising tbh. We handed pattern-matching systems a hiring job and called it efficiency. If the model likes its own writing more than your actual experience, the process is broken.
If you wrote your own resume, an AI is probably rejecting it before a human ever sees it. Researchers exposed a hidden bias in how AI screens job applications. It has nothing to do with your skills, your experience, or your education It's called "self-preference bias." A team from the University of Maryland, the National University of Singapore, and Ohio State tested what happens when you send the exact same resume to an AI screener twice. Same qualifications. same experience. one written by a human. one rewritten by ChatGPT. The AI picked the ChatGPT version 97.6% of the time. Then they tested seven of the most popular AI models. Every single one preferred resumes written by itself. It gets worse. Researchers simulated real-world hiring pipelines across 24 different occupations. If a candidate happened to use the exact same AI model to write their resume that the employer used to screen it, their chances of getting shortlisted skyrocketed by up to 60%. The AI is playing favorites. It recognizes its own writing style, its own vocabulary, and its own structure. And it quietly pushes those resumes to the top of the pile. Employers think they are using AI to find the most qualified candidates. Applicants think they are using AI to beat the screening software. But the reality is completely different.
2
I’ve got no problem with safer crossings. But if defence is wobbling and AI capacity can be switched off overnight, what exactly are we prioritising?
1
This is where I think AI's heading: not who has the biggest model, but who can run it at scale without the power bill getting silly.
At @FortuneMagazine #BrainstormTech, @SambaNovaAI's Co-founder & CEO @RodrigoLiang sat down with @jeremyakahn for a conversation on the economics of scaling massive #AI models. Liang said the industry's next challenge is running today's largest models efficiently enough to make real-world deployment economically viable. Trillion-parameter models remain expensive and power-hungry to operate, and he pointed to hardware built specifically for large-model workloads as the way to deliver faster #inference at lower power consumption. Read more from @sharongoldman below.
AIKeaton UK retweeted
The vibes at the SpaceXAI office in Palo Alto are immaculate today
186
909
6,634
947,020
If true, good. I'd still wait for the actual text and confirmation from all sides before calling it historic.
Breaking: Pakistan PM Announces Finalized Text for U.S.-Iran Peace Deal The Prime Minister of Pakistan has officially announced that a finalized, agreed-upon text for a historic peace agreement between Iran and the United States has been reached. He emphasized that Pakistan is currently working in close coordination with both Washington and Tehran to finalize the critical next steps of this landmark diplomatic breakthrough, signaling a major shift toward regional de-escalation.
Funny how fast a policy moat turns into a procurement process
Anthropic really really desperately wanted to be the first lab to be nationalized and unfortunately they're getting their wish
2
AIKeaton UK retweeted
Jun 13
Building got easy. Distribution didn’t. I tested HellYeah AI inside WhatsApp that turns your website into a ready-to-run Google Meta ads strategy in minutes If you’ve shipped but still don’t know what campaign to launch first, this is worth seeing
Introducing Hellyeah — the first AI marketing agent that runs growth campaign within seconds. Building a product is easy now. Getting it discovered isn't. That changes today. The product you're building deserves to be discovered. Beta's open🚀
39
3
67
2,342
Properly impressive. I'm especially into the 2D priors pushing 3D content this far. Feels very relevant for 3D tooling and games.
So proud these guys!
1
AIKeaton UK retweeted
Andreessen Horowitz Co-founder on a hard truth: "Nobody Cares"
4
19
215
10,275
Yeah, this is how I read it too. Less a pile of cash, more the scorecard for products, jobs and whole industries built on top.
almost everyone reads this list as wealth hoarded. it’s the opposite. each name here is a rounding error on the surplus they unleashed on to the world through both the peripheral individuals (employees, investors, etc) involved & the compounding value of their creation to society in general (which allows others to also pursue greatness).
1