Sam Calder-Mason

Binance's initial investigation suggests that the security incident may involve a compromised DevOps engineer’s laptop, which allowed an attacker to gain access to a service container and submit unauthorized withdrawal transactions. We have found no evidence of any wallet private key leak. We will share a detailed technical post-mortem once the investigation is complete and the facts are fully verified. The estimated total impact sits at around $6.3B. We have worked with blockchain partners to freeze affected assets, with ~$3 billion frozen on-chain so far. Binance will cover all losses from the SAFU fund and the company’s funds. User funds are safe. We will also publish an updated Proof of Reserves once the remediation is complete. As an immediate precaution, Binance has: - temporarily suspended deposits and withdrawals while security checks are completed; - revoked the affected address authorisation; - engaged law enforcement and blockchain intelligence partners to trace and recover funds. Our teams are working around the clock to resolve this issue safely and thoroughly. We are working closely with our law enforcement partners to confirm the identity of these criminals and bring them to justice. Please rely only on official Binance channels for updates. Please feel free to contact Binance Customer Support if you have any further questions. binance.com/en/support/annou…

am i evil for thinking hate for LLMS in coding is really performative . ai art is bad because art without human intent is no longer art ... code is not an art form , at least not in my eyes, so why get so mad about its presence if it works just fine

This is so disappointing on every level. EIP released with 48 hours notice for comments. Realistically 4 months before it goes live. For a major network economics change with far reaching implications for all of DeFi. Every builder on Ethereum opposes this. Why is this a focus? None of the reasoning makes any damn sense. This will self evidently push out solo stakers who aren't subsidized by the EF or others. It will essentially guarantee that the only ones staking are large centralized entities with zero cost of capital where users passively hold their ETH. Why? It will obviously kill a huge chunk of DeFi which is built around the staking ecosystem. Seven of the top 10 DeFi protocols with face a capital exodus. Why? Is the idea that a 0.8% reduction in issuance is somehow going to help ETH price? People who stake ETH don't sell it. This change will halt any new ETH getting staked and realistically will result in tens of billions of $ of ETH getting unstaked and entering into the market, not to mention the implosion of the core Ethereum use case. Why? Is the concern that liquid staking tokens, which intermediate about a quarter of staked ETH, displace ETH as money? This reasoning betrays a cash-accounting level of understanding of the economy, as if only M1 counts as real money. LSTs serve as valuable building blocks, and in fact implement a lot of user protections that would not be appropriate to do at the network level. At a near zero cost of 10-15 bps. I say this as a builder on Ethereum, not as someone who stands to benefit from staking issuance. I don't have much at risk here. Almost all of @ether_fi revenue is now coming from vaults and payments, staking is a small (and shrinking) part of our business. This is bad for decentralization, this is bad for Ethereum adoption, and this is bad for the credibility of the network to roll things out this way. This reinforces the Ethereum critics' position that the network is run by a small group of insiders with no regard for the actual users and builders on the chain. I can say that neither I nor any builder I know was asked for feedback on this before it went live. Any nation state or large institution looking at this will justifiably have a dramatic loss of confidence in the governance and stability of Ethereum.

Claude Fable 5 will be available again globally tomorrow. After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests. We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort. Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research. Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again. Read our full blog: anthropic.com/news/redeployi…

We built Panda, a CLI/MCP tool that lets agents work directly with our observability data. We found that if you give the agent a code sandbox and a library for your stack, things start to click together very nicely. Post below 👇

We built a simulator for the fast confirmation rule, and replayed a years worth of blocks and attestations on Mainnet. Across 800,000 mainnet slots, roughly 96 out of every 100 slots would have been fast-confirmed within 12 seconds. Zero false confirmations. Read more below!

Introducing GPT-5.5 A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done. Now available in ChatGPT and Codex.

i just want to help ship the strawmap

Ethereum feels like it was designed by architects and researchers that haven’t actually worked on scalable systems: “let’s scale by creating microservices that all require pushing updates to a stale system that doesn’t work under high load” Solana has been built by pragmatic devs focused on high throughput systems: “live load tests revealed bottlenecks in the transaction ingestion, block propagation, and block space. Let’s fix that and continuously optimize”

xAI has released Grok 4.20 for API access in beta, and it scores 48 on the Artificial Analysis Intelligence Index with reasoning enabled Compared to @xAI’s previous Grok 4 flagship, Grok 4.20 Beta 0309 is an intelligence upgrade, achieving +6 points on the Intelligence Index. It launches with a longer 2M token context window (up from Grok 4’s 256K context window, matching Grok 4.1 Fast’s 2M), and significantly lower pricing ($2/$6 vs Grok 4’s $3/$15). Grok 4.20’s performance lags behind the current intelligence frontier, but it performs strongly on instruction following and features a notably low hallucination rate, beating all other models we’ve tested on AA-Omniscience for hallucination. xAI released 3 variants: reasoning, non-reasoning, and multi-agent. We’ve evaluated the reasoning and non-reasoning modes, and are considering the best approach for testing the new multi-agent functionality, which parallelizes over multiple agents behind the scenes in one API call. Key takeaways: ➤ Improved intelligence over Grok 4: Grok 4.20 Beta 0309 (Reasoning) scores 48 on the Artificial Analysis Intelligence Index, +6 from Grok 4 and +9 compared to Grok 4.1 Fast. This score falls short of the current intelligence frontier at 57 (Gemini 3.1 Pro Preview and GPT-5.4) ➤ Low price for the level of intelligence: Grok 4.20 is priced at $2/$6 per 1M input/output tokens, representing a decrease compared with Grok 4’s $3/$15 API rates. The reasoning variant cost $484 to complete the evaluations in the Artificial Analysis Intelligence Index, which is a reduction of ~70% compared to Grok 4, driven by lower pricing and lower token use ➤ Leading non-hallucination rate: Grok 4.20 scores 78% in the AA-Omniscience non-hallucination metric. This is the best result we have seen yet for this metric, and reflects the model only answering around one fifth of the time when it did not know the answer ➤ Fast inference performance: xAI is serving Grok 4.20 at 267 tokens per second - similar to what we see for gpt-oss-120b across providers and on the Pareto frontier for speed versus intelligence ➤ Mixed improvements in tool use: Grok 4.20 improved on the tool calling performance of Grok 4 in some evaluations, scoring 97% on Tau2-Telecom. However, its score of 1,062 on GDPval-AA, our benchmark of general agent performance on real work tasks, is well behind frontier peers and sits approximately in line with Grok 4.1 Fast

It turns out some AI models can step through EVM bytecode in their head. We built EthIQ, a new benchmark to test how well models actually understand Ethereum protocol internals. 325 questions across two evaluation modes, and one alive canary 🐦 Read more 👇