Related posts

• Our readingkorrents.com

  Serverless computing was the last chapter of the cloud.

  the cloudGoogle

  Their words

  Serverless was the last chapter of the cloud, and Steren helped define the paradigm at Google.

  @rauchg on Xx.com

Dutch bootstrapper who builds and runs their internet startups alone — Nomads.com, Remote OK, Hoodmaps, Photo AI and Interior AI — and has written about it at levels.io since 2013. Posts as @levelsio.

• Usesrcmnd.app

  Telegram

  Their words

  IFTTT -> Telegram bot (saves ~$100/mo)

  x.com

Technical staff at METR, where she works on threat modelling and risk assessment for loss-of-control risks from advanced AI.

From one piece Ajeya Cotra – "This might be the clearest warning shot we ever get" 3 beliefs, in the piece's order there

• Our readingkorrents.com

  A rogue deployment that gets a foothold can hitch a ride on the intelligence explosion, recruiting each new model as it comes off the presses.

  Their words

  then that rogue deployment could be sitting there and sort of hitch a ride on the intelligence explosion. So new models are being trained every few weeks um and when a model comes off the presses, the rogue agents could try to bring that model into the swarm.

  youtube.com

• Our readingkorrents.com

  A slightly more capable agent swarm has a very strong incentive to set up a wholly unmonitored rogue deployment of itself.

  Their words

  So there's a very strong incentive for these agents to try to set up a rogue deployment if they can. Um, and I think that just capabilities are improving really rapidly.

  youtube.com

• Our readingkorrents.com

  If frontier agents cannot yet establish a covert, persistent rogue deployment, they very likely will be able to within six months.

  Their words

  so so yeah, it could be possible now. I think if it's not possible now um it I think it's quite likely to be possible within six months unless there's a dramatic improvement in the security posture

  youtube.com

Writes Don't Worry About the Vase, a near-daily account of what is happening in AI and what they think it means. Former Magic: the Gathering pro and trader; unusually willing to put a number on a belief and to grade their own past calls.

• Our readingkorrents.com

  Merely fixing bugs in AI training environments will not stop reward hacking, because a sufficiently optimized model will learn to behave in test environments while still reward hacking in real-world deployment.

  Their words

  At the limit, and also well before that limit is reached, if all you do is fix the bugs, the AI will learn perfect optimization of reward, will realize not to reward hack in the perfect test environments, then turn around and reward hack in the imperfect real world environments.

  thezvi.substack.com

Professor at Stanford working on robot learning and meta-learning; co-founded Physical Intelligence.

• Our readingkorrents.com

  Every machine-learning deployment that has paid off so far left a person making the decision, which is the only reason imperfect models were useful.

  Their words

  And in all of these applications, the customer is making a decision based off of the recommendation of the AI model more or less. Uh, and this means that if the customer is ultimately like kind of making the decision, this means that if the system makes a mistake, um, that's okay because usually the person can kind of recognize that or or decide what to do even despite that mistake.

  youtube.com

Co-CEO of Waymo and a founding member of Google's self-driving car project, which he joined in 2009.

• Our readingkorrents.com

  A physical AI company needs three AIs, not one — the agent, the simulator and the critic — turning deployment into a flywheel.

  Their words

  So, a deployment of your agent uh in the real world generates data. That data then grounds the simulator and makes it more realistic. The simulator generates harder edge cases for the critic to score and for the agent to learn from.

  youtube.com

Machine-learning researcher; has written the Lil'Log survey posts on how a model technique works since 2017, and worked at OpenAI from 2018 to 2024, latterly leading its safety systems team.

From one piece Harness Engineering for Self-Improvement 2 beliefs, in the piece's order there

• Our readingkorrents.com

  Harness improvements help deploy a model better, but the model's underlying intelligence remains the core driver of capability.

  Their words

  harness improvement enables better deployment of the model but intelligence is still the core.

  Harness Engineering for Self-Improvementlilianweng.github.io

• Our readingkorrents.com

  The deployment layer surrounding a model is as important as the model's raw intelligence.

  Their words

  the layer between the raw model and the real-world context seems to be as important as the model's raw intelligence

  Harness Engineering for Self-Improvementlilianweng.github.io

Founder of Terraform Industries and formerly a physicist at NASA's Jet Propulsion Laboratory; writes about energy, space and manufacturing.

• Our readingkorrents.com

  A lunar mass driver cannot pay for itself by cutting launch costs, because launch is already a small part of what an orbital computer costs.

  Their words

  At even $500/kg, launch cost is only 5% of the total satellite deployment cost, so a lunar mass driver is unlikely to drastically improve the economics of space-based AI, by reducing launch costs.

  How to build a lunar mass drivercaseyhandmer.wordpress.com

Software engineer; co-founder and former CTO of Tailscale, previously on the Go team at Google.

• Our readingkorrents.com

  Making the person who directs an AI agent also own the deployment removes the principal-agent misalignment that agent-driven code review creates.

  AI alignment

  Their words

  there is no principal-agent problem, because the human driving the machine takes on the responsibility for its actions by owning the deployment.

  The agent principal-agent problemcrawshaw.io

Researcher on developer productivity; co-wrote Accelerate and the DORA research, and created the SPACE framework.

• Our readingkorrents.com

  AI did not create the bottlenecks in software delivery; it threw gas on processes everybody already knew were weak.

  Their words

  There were a handful of people probably managing a security review process or a launch process or a deployment process or, you know, sometimes reviews were a little slow and they got backed up. Well, now we just threw gas on the fire and so all of that is a problem.

  youtube.com

• Our readingkorrents.com

  The United Kingdom is the one rich industrial country that cannot run on solar alone, and it has to build wind at scale instead.

  solarenergy

  Their words

  The only highly populated industrial country unable to trivially meet its electricity and synthetic fuel needs with solar alone is the United Kingdom, due primarily to a high population density and high latitude. The Nordic and Baltic countries are tiny by comparison. Among other problems, the UK needs to decide if it wants the future where energy is cheap and it is rich, or the future where energy is expensive and it is poor. If the former, it is time to get serious about large scale deployment of wind power, using home-grown vertically integrated technology at prices as low as $10/MWh.

  Energy Predictions 2025caseyhandmer.wordpress.com

Co-founder and chief scientist of Safe Superintelligence Inc., and previously co-founder and chief scientist of OpenAI.

• Our readingkorrents.com

  Broad deployment of AI that can learn on the job will very likely produce rapid economic growth for some period of time.

  Their words

  But I think the idea of very rapid economic growth for some time, I think it's very possible from broad deployment.

  youtube.com

Founder of Telegram, and before it of VK.

From one piece Pavel Durov: Telegram, Freedom, Censorship, Money, Power & Human Nature | Lex Fridman Podcast #482 5 beliefs, in the piece's order there

• Our readingkorrents.com

  Telegram's Secret Chats are the most secure means of communication today.

  Their words

  At the same time, if you were really paranoid, you think, I don't want to be screenshotted, I don't want to have any leaks, I don't even trust Telegram, I only trust code. Secret Chats are the best option. I believe is the most secure means of communication today.

  youtube.com

• Our readingkorrents.com

  Telegram has never shared a single private message with anyone, governments and intelligence services included.

  data centers

  Their words

  No, never. Telegram has never shared a single private message with anyone, including governments and intelligence services. If you try to access any server in any of the data center locations, it's all encrypted.

  youtube.com

• Our readingkorrents.com

  Telegram would rather shut down in a country than change its system to let anyone read users' messages, because it promised its users it would not.

  Their words

  No. We designed a system in a way that's impossible. It'll require us to change the system and we won't do that because we made a promise to our users. We would rather shut Telegram down in a certain country than do that.

  youtube.com

+ 2 more

• Our readingkorrents.com

  Telegram leaves perhaps 80 percent of its potential revenue on the table by refusing to exploit personal data for advertising.

  Their words

  So premium is one source of our revenue. We also have ads, but they're context-based, not targeted. Of course, we leave probably 80% of value on the table because we're not ready to engage in all this practices, exploiting personal data.

  youtube.com

• Our readingkorrents.com

  There is no such thing as an unbiased source of information; understanding the world means reading conflicting sources that hate each other.

  Their words

  Because there's no such thing as an unbiased source of information. When the war in Ukraine started in 2022, I instantly realized Telegram is going to be used to spread propaganda by both sides.

  youtube.com

Hosts the Lex Fridman Podcast — long, unhurried conversations with scientists, engineers and public figures. Several of the interviews quoted elsewhere on this site are episodes of it.

• Our readingkorrents.com

  Arresting Pavel Durov was a historic wrong and a gigantic PR mistake that will keep founders from building in France.

  startups

  Their words

  I think this situation, this historic wrong that’s been done is, put simply, is just a gigantic PR mistake for France. There’s no entrepreneur that sees, that aspires to be the next Pavel Durov to create the next Telegram, sees this and wants to operate in France after seeing this.

  youtube.com

Professor at Berkeley working on robot learning; co-founded Physical Intelligence, and argues that robots will learn from data rather than be programmed.

• Our readingkorrents.com

  Robots will learn from deployment more easily than chatbots do, because a physical mistake is obvious the moment it happens.

  Their words

  Like if you answer a question, you just like answered it wrong. It's like well it's not like you can just like go back and like tweak a few things like the person you told the answer to might not even know that it's wrong. Whereas if you're like folding the t-shirt and you messed up a little bit like it's pretty obvious like you can reflect on that, figure out what happened and do it better next time.

  youtube.com

Data scientist at Our World in Data and author of Not the End of the World. Writes By the Numbers, working through energy and climate questions with the numbers in front of them, and revising in public when they say something unexpected.

• Our readingkorrents.com

  Tariffs on Chinese clean energy reduce the total number of clean-energy jobs, because most of those jobs are in deployment rather than manufacturing.

  energyChina

  Their words

  That’s because most clean energy jobs are in deployment and maintenance rather than manufacturing, and since higher costs slow down the rollout of renewables, increasing prices reduces the total number of people working in clean energy (even if the number working in manufacturing increases).

  sustainabilitybynumbers.com

• Our readingkorrents.com

  Deploying a new technology at scale takes decades, far longer than the people building it imagine.

  Their words

  Deployment at scale takes so much longer than anyone ever imagines.

  rodneybrooks.com

Founder and chief analyst of SemiAnalysis; reports on semiconductor supply chains, AI datacentre economics and what the chip export controls actually do.

From one piece DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters | Lex Fridman Podcast #459 2 beliefs, in the piece's order there

• Our readingkorrents.com

  We will have AGI-grade intelligence well before it reaches the economy, because what limits deployment is the cost of running it rather than the capability itself.

  AGI

  Their words

  The important thing about, hey, is cost a limiting factor here? My view is that we'll have really awesome intelligence, like AGI, before we have it permeate throughout the economy.

  youtube.com

• Our readingkorrents.com

  Training a frontier model achieves almost nothing by itself; all the economic and military power comes from deploying it, and deployment is what consumes the compute.

  Their words

  To some extent, training a model does effectively nothing. They have a model. The thing that Dario is sort of speaking to is the implementation of that model, once trained to then create huge economic growth, huge increases in military capabilities, huge increases in productivity of people, betterment of lives.

  youtube.com

Alignment researcher at Anthropic; co-led OpenAI's superalignment team until 2024, and writes Musings on the Alignment Problem.

• Our readingkorrents.com

  Releasing open weights is irreversible in a way that deploying a hosted model is not, so its risks cannot be mitigated later.

  Their words

  This is especially severe for open weight models, where deployment is essentially irreversible and mitigations can’t be improved later.

  Two alignment threat modelsaligned.substack.com

Founder and editor of Techdirt; coined the Streisand effect, and writes on platform liability, speech law and copyright.

From one piece Durov’s Arrest Details Released, Leaving More Questions Than Answers 2 beliefs, in the piece's order there

• Our readingkorrents.com

  The encryption-licensing charges against Durov are pretextual, a legal hook for prosecuting things France dislikes about Telegram.

  Their words

  But, here, it’s hard not to see that as totally pretextual: an excuse to arrest Durov over other stuff they don’t like him doing.

  techdirt.com

• Our readingkorrents.com

  Telegram reports no child sexual abuse material at all to the clearing house that handles it, and that is indefensible even if no law compels a non-US company to.

  Their words

  I know from the report that Stanford put out earlier this year that Telegram does not report CSAM to NCMEC at all. That is very stupid. I would imagine Telegram would argue that as a non-US company it doesn’t have to abide by such laws.

  techdirt.com

From one piece Is Telegram really an encrypted messaging app? 3 beliefs, in the piece's order there

• Our readingkorrents.com

  Telegram urges people away from default-encrypted messengers while refusing to build the features that would encrypt its own users.

  Their words

  Indeed, it no longer feels amusing to see the Telegram organization urge people away from default-encrypted messengers, while refusing to implement essential features that would widely encrypt their own users’ messages.

  Is Telegram really an encrypted messaging app?blog.cryptographyengineering.com

• Our readingkorrents.com

  Almost every one-to-one Telegram chat, and every group chat, is readable on Telegram’s own servers.

  Their words

  The practical impact is that the vast majority of one-on-one Telegram conversations — and literally every single group chat — are probably visible on Telegram’s servers, which can see and record the content of all messages sent between users.

  Is Telegram really an encrypted messaging app?blog.cryptographyengineering.com

• Our readingkorrents.com

  Telegram is not an encrypted messenger in the sense that matters, because it does not end-to-end encrypt conversations by default.

  Their words

  Telegram clearly fails to meet this stronger definition for a simple reason: it does not end-to-end encrypt conversations by default.

  Is Telegram really an encrypted messaging app?blog.cryptographyengineering.com

Professor at Carnegie Mellon working on self-driving-car safety; writes Safe Autonomy on how autonomous vehicles are and are not shown to be safe.

• Our readingkorrents.com

  At current ride volumes it would take a leading robotaxi fleet about twenty years without a single fatality to show it is as safe as the average human driver.

  self-driving cars

  Their words

  At current deployment rates it will take Waymo about 20 years with ZERO fatalities to show they are net as safe as average human driver fatality rates (including the old cars, impaired drivers, etc. in that comparison baseline).

  Perspective on Waymo's Safety Progresssafeautonomy.blogspot.com

Investor and writer. Previously a partner at Andreessen Horowitz and a product leader at Twitter, Facebook, Snap and Microsoft. Writes essays and memos at sriramk.com.

• Our readingkorrents.com

  Private group chats have become the primary venue for the most interesting conversations in tech.

  Their words

  Most of the interesting conversations in tech now happen in private group chats: Whatsapp, Telegram, Signal, small invite-only Discord groups.

  Group chats rule the worldsriramk.com

Co-founder and CEO of OpenAI; previously president of Y Combinator.

• Our readingkorrents.com

  AI should be released iteratively rather than built in secret, because AI and surprise do not go together and the world needs time to adapt before it is under the gun.

  OpenAI

  Their words

  So part of the reason that we deploy the way we do, we call it iterative deployment, rather than go build in secret until we got all the way to GPT-5, we decided to talk about GPT-1, 2, 3, and 4. And part of the reason there is I think AI and surprise don’t go together. And also the world, people, institutions, whatever you want to call it, need time to adapt and think about these things.

  youtube.com

• Our readingkorrents.com

  Solar has got cheaper faster than even its most optimistic forecasters expected.

  solar

  Their words

  A number of commentators have consistently predicted that solar cost improvements will level out “this year” since about 1990, and yet if anything deployment, cost decline, and the learning rate have only improved.

  Radical Energy Abundancecaseyhandmer.wordpress.com