Andres Antillon

[Tibo @thsottiaux]

Morning. The last 48 hours of Codex and ChatGPT Work have been intense! Three important updates:

- Temporarily removing the 5 hour usage limit restriction for all Plus, Business and Pro plans
- Rolling out changes that will make GPT 5.6 Sol more efficient across the board and that will be reflected in less usage being used so that it can take you further. Exact impact to be quantified and shared
- We hit 6M active users, and are landing a usage reset in the next hour

Go do things

[Andres Antillon @axantillon]

@willyhopps https://t.co/ekF9F0dE7d

[Andres Antillon @axantillon]

@willyhopps 🫶

[Andres Antillon @axantillon]

@rshanreddy @JoinEdyfi @justoutquan @rtwlz @willyhopps @_jzhao @McDonalds looooong way 🫡 proud to have shared the stage with you all at one point

[Justin Quan @justoutquan]

It's been 3 years since AI exploded, but nobody's life has improved outside of Silicon Valley.

Tomo is for real humans, like my little brother Lucas

and 10,000 other humans taking a chance on themselves every single day.

[Justin Quan @justoutquan]

Why you should never bet on yourself.

Tomo, launching june 25th

[Will Hoppin @willyhopps]

taste is inherently human and subjective - AI should only help you execute yours with minimal friction

[Nayib Bukele @nayibbukele]

The key to saving the environment is not looking backward, it’s moving forward.

I realized this the first time I visited Italy twenty years ago. Everything was clean and green. The rivers sparkled. The lesson for me was obvious: the answer is not underdevelopment. The answer is progress.

When China was poor, the air was so polluted that people could barely see the blue sky. Today, blue skies have returned to their cities. Development does not only create wealth, it also provides the resources needed to restore and protect the environment.

Some environmentalists want us to preserve every aspect of our biodiversity, including the mosquitoes for example, so that researchers can fly in once every ten years from their universities (which build particle accelerators and billion-dollar laboratories with their pocket money), study our ecosystems, and count how many people died from dengue outbreaks.

They want to buy our air through carbon credits. If carbon credits were such a great deal, they would be selling them to us, not the other way around.

Cleaning every river, lake, and water source in El Salvador, and ensuring they remain clean and sparkling, would cost roughly $12 billion. Where is that money supposed to come from without economic development? Carbon credits?

The path forward for our country is the path of Japan and Singapore, not the path of the Congo.

[vitalik.eth @VitalikButerin]

Crypto privacy is needed if you want to make API calls without compromising the information of your access patterns.

eg. even with a local AI agent, you can learn a lot about what someone is doing if you see all of their search engine calls

first-order solution to that is to make those calls through mixnet

but then (or in fact, even without the mixnet) the providers will get DoSed, and they will demand an anti-DoS mechanism, and realistically payment per call

by default that will be credit card or some corposlop "yeah we'll get to the privacy later" stablecoin thing

so we need crypto privacy

But yes, for privacy you have to think full stack. Local AI agent layer is very important.

It is like longevity: if there are 10 things damaging your body, curing one of them increases your longevity by 11%, curing two by 25%, and curing three by 42% (1 / (1 - 0.3) minus 100% base). Risks from data leakage are similar, and so mitigations similarly compound super-additively.

[Ethan Lipnik @EthanLipnik]

Lowkey I’m already building this

[ashton kutcher @aplusk]

If you liked LLM wrapper companies you’re gonna love agent wrapper companies😂🦞

[Riley Walz @rtwlz]

EXCLUSIVE SCOOP: I joined the wonderful Labs team at OpenAI. I am learning a lot and have l enjoyed it very much.

[Balaji @balajis]

AI inside, crypto outside.

Within your tribe's trusted perimeter, you give AI all the context and let it figure it out.

But outside your tribe's trusted perimeter, you hit a wall of AI spam. Now you need cryptographic verification to see what's true.

[Andrej Karpathy @karpathy]

First there was chat, then there was code, now there is claw. Ez

[Balaji @balajis]

I have never been more bullish on crypto.

Because the rules-based order is collapsing and the code-based order is rising. So the short term price doesn’t matter.

As international law breaks down, we will need not just onchain currencies, but onchain companies. As the post-war order breaks down, we’ll similarly need the post-internet order. States will fail, and the network will take their place.

We need internet capitalism, we need internet democracy, and we need internet privacy. So we need cryptocurrency.

[Andrej Karpathy @karpathy]

A few random notes from claude coding quite a bit last few weeks.

Coding workflow. Given the latest lift in LLM coding capability, like many others I rapidly went from about 80% manual+autocomplete coding and 20% agents in November to 80% agent coding and 20% edits+touchups in December. i.e. I really am mostly programming in English now, a bit sheepishly telling the LLM what code to write... in words. It hurts the ego a bit but the power to operate over software in large "code actions" is just too net useful, especially once you adapt to it, configure it, learn to use it, and wrap your head around what it can and cannot do. This is easily the biggest change to my basic coding workflow in ~2 decades of programming and it happened over the course of a few weeks. I'd expect something similar to be happening to well into double digit percent of engineers out there, while the awareness of it in the general population feels well into low single digit percent.

IDEs/agent swarms/fallability. Both the "no need for IDE anymore" hype and the "agent swarm" hype is imo too much for right now. The models definitely still make mistakes and if you have any code you actually care about I would watch them like a hawk, in a nice large IDE on the side. The mistakes have changed a lot - they are not simple syntax errors anymore, they are subtle conceptual errors that a slightly sloppy, hasty junior dev might do. The most common category is that the models make wrong assumptions on your behalf and just run along with them without checking. They also don't manage their confusion, they don't seek clarifications, they don't surface inconsistencies, they don't present tradeoffs, they don't push back when they should, and they are still a little too sycophantic. Things get better in plan mode, but there is some need for a lightweight inline plan mode. They also really like to overcomplicate code and APIs, they bloat abstractions, they don't clean up dead code after themselves, etc. They will implement an inefficient, bloated, brittle construction over 1000 lines of code and it's up to you to be like "umm couldn't you just do this instead?" and they will be like "of course!" and immediately cut it down to 100 lines. They still sometimes change/remove comments and code they don't like or don't sufficiently understand as side effects, even if it is orthogonal to the task at hand. All of this happens despite a few simple attempts to fix it via instructions in CLAUDE . md. Despite all these issues, it is still a net huge improvement and it's very difficult to imagine going back to manual coding. TLDR everyone has their developing flow, my current is a small few CC sessions on the left in ghostty windows/tabs and an IDE on the right for viewing the code + manual edits.

Tenacity. It's so interesting to watch an agent relentlessly work at something. They never get tired, they never get demoralized, they just keep going and trying things where a person would have given up long ago to fight another day. It's a "feel the AGI" moment to watch it struggle with something for a long time just to come out victorious 30 minutes later. You realize that stamina is a core bottleneck to work and that with LLMs in hand it has been dramatically increased.

Speedups. It's not clear how to measure the "speedup" of LLM assistance. Certainly I feel net way faster at what I was going to do, but the main effect is that I do a lot more than I was going to do because 1) I can code up all kinds of things that just wouldn't have been worth coding before and 2) I can approach code that I couldn't work on before because of knowledge/skill issue. So certainly it's speedup, but it's possibly a lot more an expansion.

Leverage. LLMs are exceptionally good at looping until they meet specific goals and this is where most of the "feel the AGI" magic is to be found. Don't tell it what to do, give it success criteria and watch it go. Get it to write tests first and then pass them. Put it in the loop with a browser MCP. Write the naive algorithm that is very likely correct first, then ask it to optimize it while preserving correctness. Change your approach from imperative to declarative to get the agents looping longer and gain leverage.

Fun. I didn't anticipate that with agents programming feels *more* fun because a lot of the fill in the blanks drudgery is removed and what remains is the creative part. I also feel less blocked/stuck (which is not fun) and I experience a lot more courage because there's almost always a way to work hand in hand with it to make some positive progress. I have seen the opposite sentiment from other people too; LLM coding will split up engineers based on those who primarily liked coding and those who primarily liked building.

Atrophy. I've already noticed that I am slowly starting to atrophy my ability to write code manually. Generation (writing code) and discrimination (reading code) are different capabilities in the brain. Largely due to all the little mostly syntactic details involved in programming, you can review code just fine even if you struggle to write it.

Slopacolypse. I am bracing for 2026 as the year of the slopacolypse across all of github, substack, arxiv, X/instagram, and generally all digital media. We're also going to see a lot more AI hype productivity theater (is that even possible?), on the side of actual, real improvements.

Questions. A few of the questions on my mind:
- What happens to the "10X engineer" - the ratio of productivity between the mean and the max engineer? It's quite possible that this grows *a lot*.
- Armed with LLMs, do generalists increasingly outperform specialists? LLMs are a lot better at fill in the blanks (the micro) than grand strategy (the macro).
- What does LLM coding feel like in the future? Is it like playing StarCraft? Playing Factorio? Playing music?
- How much of society is bottlenecked by digital knowledge work?

TLDR Where does this leave us? LLM agent capabilities (Claude & Codex especially) have crossed some kind of threshold of coherence around December 2025 and caused a phase shift in software engineering and closely related. The intelligence part suddenly feels quite a bit ahead of all the rest of it - integrations (tools, knowledge), the necessity for new organizational workflows, processes, diffusion more generally. 2026 is going to be a high energy year as the industry metabolizes the new capability.

[vitalik.eth @VitalikButerin]

We need more DAOs - but different and better DAOs.

The original drive to build Ethereum was heavily inspired by decentralized autonomous organizations: systems of code and rules that lived on decentralized networks that could manage resources and direct activity, more efficiently and more robustly than traditional governments and corporations could.

Since then, the concept of DAOs has migrated to essentially referring to a treasury controlled by token holder voting - a design which "works", hence why it got copied so much, but a design which is inefficient, vulnerable to capture, and fails utterly at the goal of mitigating the weaknesses of human politics. As a result, many have become cynical about DAOs.

But we need DAOs.

* We need DAOs to create better oracles. Today, decentralized stablecoins, prediction markets, and other basic building blocks of defi are built on oracle designs that we are not satisfied with. If the oracle is token based, whales can manipulate the answer on a subjective issue and it becomes difficult to counteract them. Fundamentally, a token-based oracle cannot have a cost of attack higher than its market cap, which in turn means it cannot secure assets without extracting rent higher than the discount rate. And if the oracle uses human curation, then it's not very decentralized. The problem here is not greed. The problem is that we have bad oracle designs, we need better ones, and bootstrapping them is not just a technical problem but also a social problem.
* We need DAOs for onchain dispute resolution, a necessary component of many types of more advanced smart contract use cases (eg. insurance). This is the same type of problem as price oracles, but even more subjective, and so even harder to get right.
* We need DAOs to maintain lists. This includes: lists of applications known to be secure or not scams, lists of canonical interfaces, lists of token contract addresses, and much more.
* We need DAOs to get projects off the ground quickly. If you have a group of people, who all want something done and are willing to contribute some funds (perhaps in exchange for benefits), then how do you manage this, especially if the task is too short-duration for legal entities to be worth it?
* We need DAOs to do long-term project maintenance. If the original team of a project disappears, how can a community keep going, and how can new people coming in get the funding they need?

One framework that I use to analyze this is "convex vs concave" from https://t.co/1BrMsUAKWK . If the DAO is solving a concave problem, then it is in an environment where, if faced with two possible courses of action, a compromise is better than a coin flip. Hence, you want systems that maximize robustness by averaging (or rather, medianing) in input from many sources, and protect against capture and financial attacks. If the DAO is solving a convex problem, then you want the ability to make decisive choices and follow through on them. In this case, leaders can be good, and the job of the decentralized process should be to keep the leaders in check.

For all of this to work, we need to solve two problems: privacy, and decision fatigue. Without privacy, governance becomes a social game (see https://t.co/uMXcuzQNjM ). And if people have to make decisions every week, for the first month you see excited participation, but over time willingness to participate, and even to stay informed, declines.

I see modern technology as opening the door to a renaissance here. Specifically:

* ZK (and in some cases MPC/FHE, though these should be used only when ZK along cannot solve the problem) for privacy
* AI to solve decision fatigue
* Consensus-finding communication tools (like https://t.co/Nzord32Ub1, but going further)

AI must be used carefully: we must *not* put full-size deepseek (or worse, GPT 5.2) in charge of a DAO and call it a day. Rather, AI must be put in thoughtfully, as something that scales and enhances human intention and judgement, rather than replacing it. This could be done at DAO level (eg. see how https://t.co/pCHI7Mlo2m works), or at individual level (user-controlled local LLMs that vote on their behalf).

It is important to think about the "DAO stack" as also including the communication layer, hence the need for forums and platforms specially designed for the purpose. A multisig plus well-designed consensus-finding tools can easily beat idealized collusion-resistant quadratic funding plus crypto twitter.

But in all cases, we need new designs. Projects that need new oracles and want to build their own should see that as 50% of their job, not 10%.

Projects working on new governance designs should build with ZK and AI in mind, and they should treat the communication layer as 50% of their job, not 10%.

This is how we can ensure the decentralization and robustness of the Ethereum base layer also applies to the world that gets built on top.

[vitalik.eth @VitalikButerin]

Ethereum itself must pass the walkaway test.

Ethereum is meant to be a home for trustless and trust-minimized applications, whether in finance, governance or elsewhere. It must support applications that are more like tools - the hammer that once you buy it's yours - than like services that lose all functionality once the vendor loses interest in maintaining them (or worse, gets hacked or becomes value-extractive). Even when applications do have functionality that depends on a vendor, Ethereum can help reduce those dependencies as much as possible, and protect the user as much as possible in those cases where the dependencies fail.

But building such applications is not possible on a base layer which itself depends on ongoing updates from a vendor in order to continue being usable - even if that "vendor" is the all core devs process. Ethereum the blockchain must have the traits that we strive for in Ethereum's applications. Hence, Ethereum itself must pass the walkaway test.

This means that Ethereum must get to a place where we _can ossify if we want to_. We do not have to stop making changes to the protocol, but we must get to a place where Ethereum's value proposition does not strictly depend on any features that are not in the protocol already.

This includes the following:

* Full quantum-resistance. We should resist the trap of saying "let's delay quantum-resistance until the last possible moment in the name of ekeing out more efficiencies for a while longer". Individual users have that right, but the protocol should not. Being able to say "Ethereum's protocol, as it stands today, is cryptographically safe for a hundred years" is something we should strive to get to as soon as possible, and insist on as a point of pride.
* An architecture that can expand to sufficient scalability. The protocol needs to have the properties that allow it to expand to many thousands of TPS over time, most notably ZK-EVM validation and data sampling through PeerDAS. Ideally, we get to a point where further scaling is done through "parameter only" changes - and ideally _those_ changes are not BPO-style forks, but rather are made with the same validator voting mechanism we use for the gas limit.
* A state architecture that can last decades. This means deciding, and implementing, whatever form of partial statelessness and state expiry will let us feel comfortable letting Ethereum run with thousands of TPS for decades, without breaking sync or hard disk or I/O requirements. It also means future-proofing the tree and storage types to work well with this long-term environment.
* An account model that is general-purpose (this is "full account abstraction": move away from enshrined ECDSA for signature validation)
* A gas schedule that we are confident is free of DoS vulnerabilities, both for execution and for ZK-proving
* A PoS economic model that, with all we have learned over the past half decade of proof of stake in Ethereum and full decade beyond, we are confident can last and remain decentralized for decades, and supports the usefulness of ETH as trustless collateral (eg. in governance-minimized ETH-backed stablecoins)
* A block building model that we are confident will resist centralization pressure and guarantee censorship resistance even in unknown future environments

Ideally, we do the hard work over the next few years, to get to a point where in the future almost all future innovation can happen through client optimization, and get reflected in the protocol through parameter changes. Every year, we should tick off at least one of these boxes, and ideally multiple. Do the right thing once, based on knowledge of what is truly the right thing (and not compromise halfway fixes), and maximize Ethereum's technological and social robustness for the long term.

Ethereum goes hard.

This is the gwei.

[Amjad Masad @amasad]

I wrote this manifesto many years ago on what happens when programming becomes more accessible. It’s maybe idealistic, but still interesting to look back since we’ve basically arrived and I believe the world will change rapidly from here:

- - -

Computers are the most powerful tools to exist in the history of humanity. Sadly most people are mere consumers of these machines. Only a relative few -- the professional software developer -- can use this superpower to its fullest extent. It’s creating an unbalanced world where there are programmers, and then there are those who are programmed. The situation is analogous to literacy before the printing press, where only the powerful had access to books and written communication. The invention of the printing press led to democratic, scientific, and industrial revolutions, but it took about a century from creation to revolution, which meant that an entire generation of people to grow up with new skills, outlooks, and ways of learning and communicating.

When there is widespread code literacy, we believe that it will transform humanity again and perhaps in more radical ways. It's hard to imagine what the world will look like when there is a generation of people who grew up with computer superpowers. But we can be sure about a few things:

•Technological power will be decentralized: When everyone can code, there will be less reliance on power centers to manage all information technology. Power centers like Silicon Valley will decentralize into internet communities.

•Wealth will be more accessible: When people have the tools to create and participate in the global digital economy, they will have the ability to lift themselves out of poverty. The wealthy and the poor will not differ by pidegree, instead anyone who can generate good ideas can be wealthy.

•Software will be more local: No one knows better what a community needs other than the community itself. Instead of a handful of giant companies creating the software for everyone in the world, we'll see more personal and community-grown software.

•Education will be transformed: When people can be economically productive much earlier in life, education systems will change to become more tailored to individuals and more pragmatic. Since the best way to learn is to create, people will become more efficient learners by building computer simulations.

•Institutions will be more democratic: Totalitarian regimes have relied on technological supremacy since the beginning of time to subject people. But when people have agency over technology, the power balance will improve.

Mission

Replit will bring the next billion software creators online, and will accelerate the shift towards a world where software and the internet are truly The Great Equalizer. Soon anyone, regardless of place or economic status, will be able to create software and generate wealth on the internet. If we’re successful, anyone who's willing to learn and generate good ideas will be wealthy.

[Riley Walz @rtwlz]

We cloned Google Photos, except it's Epstein's camera roll