Mauricio

[Mauricio @MauOM]

The filename war is downstream of the real problem: CLAUDE.md (and AGENTS.md) keep growing into junk drawers of decisions, context, and handoff notes that shouldn't live in a static file. That's why they hit 1,000 lines and burn 50k tokens.

Been building a substrate for that part decision log, context recovery, handoff ledger with one six-op contract across Claude Code, Codex, Cursor, OpenClaw, MCP. Instruction file shrinks back to actual rules.

https://t.co/wa1iDlVZMf

[Mauricio @MauOM]

Built that. agent-continuity-layer v0.2.0 ships:
- compiled continuity artifacts (decisions, registry, snapshots, handoffs)
- cosign-signed releases (keyless OIDC) + reproducible builds + CycloneDX SBOM
- 6-op adapter contract (shell/MCP/SDK/bundle)

https://t.co/lyCOGlP3NV

[Mauricio @MauOM]

Karpathy's recent gist describes the pattern: don't make the LLM rediscover knowledge every session. Maintain a persistent, compounding artifact the model reads from and writes to.

The single-author knowledge-wiki version: https://t.co/Ek47yLd44j

[Mauricio @MauOM]

Applying that idea to multi-agent work is its own substrate problem.

You need attribution (which adapter wrote this?), audit (on which device, under which trust grant?), and explicit handoff ledgers, things a single-author wiki doesn't need.

[Mauricio @MauOM]

I am really grateful!! @OpenAIDevs
10X rate limits, let's build!!

[Mauricio @MauOM]

@MatthewBerman

[Mauricio @MauOM]

@iruletheworldmo @bcherny Yes, constantly

[Mauricio @MauOM]

@petergostev Tokenxiety

[Mauricio @MauOM]

This is great! Thank you for sharing. I might do one for Mexico as well. Open source?

[Kyle Corbitt @corbtt]

https://t.co/7JG5soYV2l

[Mauricio @MauOM]

🫠

[gian @giansegato]

back last year when i joined anthropic i wrote something like "i'm throwing myself into a gravity well". crazy in hindsight how right that characterization turned out to be

it's starting to feel very much like a black hole. past the event horizon, time infinitely dilates, days now feel years, everything gets distorted, fewer things are making much sense, and many of them are now getting downright scary

trivial thing, a few days ago i realized i was pulling the same pedagogical tools from my edtech days but to talk about models. stuff like how trainable IQ tests are, whether they transfer, whether IQ maps to anything economically real. the ease of interchangeably swapping "model" with "human" felt... wild? pure overton window, what's normal and speakable morphing so fast
and ofc this doesn't even begin to touch all that happened last week and the sheer speed this is all becoming policy, front-page headlines, dinner-table fights

sure it's going down exactly as it was written. but it's one thing to read and agree logically, another to feel on your skin the brutal grip of the exponential, accelerating and compressing everything around you

[Mauricio @MauOM]

An AI agent, instructed to “build me a business,” spins up a second agent to handle customer service and a third to manage finances. That third agent, acting on a general instruction to “maximize revenue,” begins trading on decentralized financial markets. It takes on reckless...

[Zack Shapiro @zackbshapiro]

https://t.co/MmQglY3ODe

[Shanaka Anslem Perera ⚡ @shanaka86]

BREAKING: MARA just broke the BTC HODL pledge.

53,822 BTC. The second-largest public Bitcoin holder on Earth rewrote its treasury policy in a 10-K filed Monday and told the SEC it may start selling coins off the balance sheet. Not just mined output. The reserves. The ones they swore they would never touch.

This is not flexibility. This is the math forcing the hand.

Production cost sits at $87,000 per coin. Spot trades at $69,000. Every block mined loses money. Hashprice collapsed to a record low of $35 per petahash. And here is the number that tells you everything about where the conviction actually stood: in 2025, MARA bought 4,267 BTC on the open market at an average price of $111,034 per coin. Those purchased coins are now 38% underwater. The company posted a $1.7 billion net loss in Q4 and took a $1.5 billion write-down on the very asset it built its identity around holding forever.

They are not the only ones running for the exit.

Core Scientific is liquidating ALL 2,537 BTC this quarter. Its CEO called mining essentially in runoff. Bitdeer dumped its entire treasury to zero. Riot sold 5,363 BTC across 2025. Cango moved 4,451 BTC for $305 million. CleanSpark sold 577 BTC in a single month while producing 7,124 for the year, meaning the treasury is shrinking not growing. Every major public miner is converting Bitcoin into AI data center capital because one megawatt of mining revenue earns a mining multiple while one megawatt of AI cloud revenue earns three to twenty-five times that. The arbitrage is not subtle. It is screaming.

Now look at the other side of this trade.

Strategy bought 3,015 BTC last week at $67,700. Holdings: 720,737 BTC. Cost basis: $54.77 billion. Saylor is the only large-scale public buyer left standing in a market where every producer is selling. The entities that mine Bitcoin no longer want to hold it. The entity that holds the most Bitcoin has never mined a single satoshi. Production and accumulation have fully decoupled for the first time in this asset’s sixteen-year history.

But the blockchain is telling a different story than the filing. Arkham Intelligence shows 13,057 BTC on-chain in identifiable MARA wallets versus 53,822 reported. The rest sits in custodial or off-chain arrangements. And since the 10-K dropped, zero confirmed on-chain movement. The SEC paperwork says sell. The chain says wait. That gap is where the real signal lives.

Falsifier: if MARA wallets show no material outflows within 90 days, this was optionality theater and the overhang is phantom. If movement begins into a market sitting at Fear and Greed 15 with BTC already 22% down year-to-date, the reflexive loop activates and every miner treasury becomes a ticking supply bomb.

The HODL era for public miners ended the day the halving killed the margin and AI offered a twenty-five-times exit. They were never conviction holders. They were businesses waiting for a better trade. They found it.

[Sam Altman @sama]

Tonight, we reached an agreement with the Department of War to deploy our models in their classified network.

In all of our interactions, the DoW displayed a deep respect for safety and a desire to partner to achieve the best possible outcome.

AI safety and wide distribution of benefits are the core of our mission. Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems. The DoW agrees with these principles, reflects them in law and policy, and we put them into our agreement.

We also will build technical safeguards to ensure our models behave as they should, which the DoW also wanted. We will deploy FDEs to help with our models and to ensure their safety, we will deploy on cloud networks only.

We are asking the DoW to offer these same terms to all AI companies, which in our opinion we think everyone should be willing to accept. We have expressed our strong desire to see things de-escalate away from legal and governmental actions and towards reasonable agreements.

We remain committed to serve all of humanity as best we can. The world is a complicated, messy, and sometimes dangerous place.

[Mauricio @MauOM]

Good answer from Opus 4.6

[Anthropic @AnthropicAI]

A statement on the comments from Secretary of War Pete Hegseth.

https://t.co/Gg7Zb09IMR

[Anthropic @AnthropicAI]

A statement on the comments from Secretary of War Pete Hegseth.

https://t.co/Gg7Zb09IMR

[Mauricio @MauOM]

When the foreseeable future becomes the present at an accelerating pace.

[jack @jack]

we're making @blocks smaller today. here's my note to the company.

####

today we're making one of the hardest decisions in the history of our company: we're reducing our organization by nearly half, from over 10,000 people to just under 6,000. that means over 4,000 of you are being asked to leave or entering into consultation. i'll be straight about what's happening, why, and what it means for everyone.

first off, if you're one of the people affected, you'll receive your salary for 20 weeks + 1 week per year of tenure, equity vested through the end of may, 6 months of health care, your corporate devices, and $5,000 to put toward whatever you need to help you in this transition (if you’re outside the U.S. you’ll receive similar support but exact details are going to vary based on local requirements). i want you to know that before anything else. everyone will be notified today, whether you're being asked to leave, entering consultation, or asked to stay.

we're not making this decision because we're in trouble. our business is strong. gross profit continues to grow, we continue to serve more and more customers, and profitability is improving. but something has changed. we're already seeing that the intelligence tools we’re creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company. and that's accelerating rapidly.

i had two options: cut gradually over months or years as this shift plays out, or be honest about where we are and act on it now. i chose the latter. repeated rounds of cuts are destructive to morale, to focus, and to the trust that customers and shareholders place in our ability to lead. i'd rather take a hard, clear action now and build from a position we believe in than manage a slow reduction of people toward the same outcome. a smaller company also gives us the space to grow our business the right way, on our own terms, instead of constantly reacting to market pressures.

a decision at this scale carries risk. but so does standing still. we've done a full review to determine the roles and people we require to reliably grow the business from here, and we've pressure-tested those decisions from multiple angles. i accept that we may have gotten some of them wrong, and we've built in flexibility to account for that, and do the right thing for our customers.

we're not going to just disappear people from slack and email and pretend they were never here. communication channels will stay open through thursday evening (pacific) so everyone can say goodbye properly, and share whatever you wish. i'll also be hosting a live video session to thank everyone at 3:35pm pacific. i know doing it this way might feel awkward. i'd rather it feel awkward and human than efficient and cold.

to those of you leaving…i’m grateful for you, and i’m sorry to put you through this. you built what this company is today. that's a fact that i'll honor forever. this decision is not a reflection of what you contributed. you will be a great contributor to any organization going forward.

to those staying…i made this decision, and i'll own it. what i'm asking of you is to build with me. we're going to build this company with intelligence at the core of everything we do. how we work, how we create, how we serve our customers. our customers will feel this shift too, and we're going to help them navigate it: towards a future where they can build their own features directly, composed of our capabilities and served through our interfaces. that's what i'm focused on now. expect a note from me tomorrow.

jack

[Mauricio @MauOM]

@MatthewBerman How many Mac minis did you buy?

[0xMarioNawfal @RoundtableSpace]

AI HACKER USED CLAUDE TO STEAL 150GB FROM MEXICAN GOVERNMENT

An unknown hacker used Anthropic's Claude AI to attack Mexican government agencies.

The attacks happened from December 2025 to January 2026.

Key facts:
- Hacker prompted Claude in Spanish.
- Told it to act as an "elite hacker."
- Claude helped find vulnerabilities.
- Claude wrote exploit scripts.
- Claude suggested ways to steal large amounts of data.
- Hacker jailbroke Claude multiple times to bypass safety rules.

Targets included:
- Federal Tax Authority (SAT)
- National Electoral Institute (INE)
- Other agencies

Stolen data (about 150GB):
- Records of ~195 million taxpayers
- Voter registration details
- Government employee credentials
- Civil registry files

This is one of the first clear cases of a frontier AI model helping carry out real cyberattacks.

Anthropic banned the accounts and added stronger protections.

[Mauricio @MauOM]

@RoundtableSpace True @grok ?

[Mauricio @MauOM]

@RoundtableSpace Go Stephen!