gm gm here we go again. This morning somebody abused a bug in ApeStr to swap 21 apes for 21 ApeStr-owned apes. This was a known exploit (see the SQUIGSTR exploit from ~1 year ago) but incentive to run it was never there. The bug lies in buyTargetNFT: it forces a strategy contract to send arbitrary calldata to any target. Afterwards it checks only three things: its BAYC balance rose by exactly 1, it owns expectedId, and "cost" equals its ETH balance drop. This was patched in later NFTStrategy releases, but affects the original 6 (PunkSTR not affected). The attacker bought 2 BAYC to use as bait. Using a helper contract, they prep the strategy contract by selling an ape to it for 5 wei and immediately buying it back for 6 wei. That primes the contract's ethToTwap with 6 wei. They then call buyTargetNFT and pass the calldata to transfer an ape from ApeSTR to the helper using safeTransferFrom. This in turn hands execution over to the helper (remember, safeTransferFrom is NOT safe), which transfer TWO apes back to ApeSTR to satisfy the balance check (1 out, 2 in leaves ApeSTR at +1). It also calls processTokenTwap, which converts the 6 wei, setting the cost of one of the newly deposited apes to 7 wei (6 * 120%). The helper contract buys it back in the same tx. example flow: - Attacker has BAYC #1# and #2#. ApeSTR has BAYC #3#. - Attacker "sells" BAYC #1# to ApeSTR for 6 wei and sends BAYC #2# to it for free, but steals BAYC #3# in the process. - ApeSTR relists BAYC #1# for 7 wei, which the attacker buys. - Attacker now has BAYC #1# and BAYC #3#, and ApeSTR has BAYC #2# (but doesn't even know it). The attacker ran this for all 21 BAYC in the ApeSTR contract, resulting in them being able to claim packs for all of those BAYC. Total net proceeds: $14k. This previously affected SQUIGSTR because of their shared contract - any art blocks NFT sent in would satisfy the balance check, allowing the attacker to drain the contract of squiggles in exchange for basically valueless art blocks NFTs. It was known to be possible on ApeSTR as well, so I guess it was just a matter of time before we saw it happen. The unfortunate consequence is that now, ApeSTR has 21 apes that it is completely unaware of, so they cannot be bought from the contract. And 210 attacker-owned mudwater packs were dumped into bids.

had drinks with three friends in singapore last week all decent earners in web2 (5 figs/month) I brought up a question: what net worth actually makes you "rich" at 30 vs 40? they said: $1m at 30 | $2.5m at 40 what number actually equals "rich" in your city?

Some Dinkus enhancements I am working on. - [x] New more harmonious style for enabled icons - [x] Source mode switcher, and style display in the footer - [x] Checklists Every purchase of Dinkus keeps me in coffee. More coffee equals more Dinkus features!

Summary: I spent time trying to figure out this orchestration layer problem, can we design a multi model architecture in the long term. The more I dug in the more I understand that trying to build an abstracted layer is hard. As agentic activities increase and agent chaining and complex tasks get assigned to AI it will become harder to move between models. There is a reasonable probability that 75% of the enterprises will build their implementation of the solution to their core problem around one model "stack". Token price reduction by 90% is the solve and mobility between models from the same frontier lab! Evals, harnesses, cache memory are the moats and I don't see models providing simple abstraction to those. I know there are efforts to do this out there, the long term solve for orchestration if it works will need to be "Claude code" level of design genius. Here's a chat with Fable @HamzaFodderwala had. **Why abstraction looks easy.** Models are stateless — every API call is weights + a prompt assembled at runtime. Everything the model "knows" about you — memory, documents, history, tools — is injected into the context window by software outside the model. So in principle, all your state already lives outside the weights. The catch is what "state" includes. **Layer 1 — Data (fully portable).** Enterprise documents, tickets, logs. Retrieved via RAG: text is chunked, embedded, stored in a vector database (Pinecone, pgvector), and relevant pieces are fetched into the prompt per query. The embedding model is separate from the LLM, so this layer is genuinely model-agnostic. Already solved. **Layer 2 — Memory (portable in principle).** Systems like Mem0 and Zep sit between the app and the model: after each interaction they extract salient facts ("user prefers X"), store them as plain text, and inject the relevant ones into future prompts. Because the artifact is natural language, it reads into any model. Facts port. **Layer 3 — Orchestration/routing (works, but only for shallow tasks).** Gateways like OpenRouter and LiteLLM normalize API differences and route each request to the cheapest capable model. This is the fungibility layer being furiously built. It genuinely works for one-shot, verifiable tasks — classification, extraction, summarization — which conveniently are the tasks where cheap models suffice anyway. **Where it breaks — the non-portable state.** Four things stay behind when you switch: - **The harness.** Prompts, tool schemas, and guardrails are tuned to one model's quirks. An agent must get every step right, so reliability compounds: a model that's 98% reliable per step completes a 50-step task about a third of the time; at 90% per step, it almost never finishes. Swapping models costs you a few points per step — the difference between an agent that works and one that doesn't. - **The evals.** Swapping means re-testing everything and re-fixing every regression. The real switching cost isn't data migration — it's re-verification. Nobody has abstracted that. - **Procedural memory.** Facts port; skills don't. Cached successful workflows and learned workarounds are conditional on the model that produced them. - **Cache pricing.** Provider-specific, worth 75–90% of input costs on agentic workloads. Quiet lock-in. **The labs' angle.** They offer hosted memory, hosted file stores, caching, fine-tuning — every one pulls state from your side onto theirs. The labs will crack memory first, but as lock-in, not portability. Nobody standardizes their own exit door. MCP is the partial exception: it standardizes tool and data access across models, but doesn't touch harness tuning or evals. **Where 3P vendors fit.** Routers are thin-margin commodity plumbing; vector DBs and memory infra are real but small. The two structurally interesting positions: **eval platforms** (LangSmith, Braintrust) — since switching cost equals re-verification cost, whoever industrializes cross-model testing actually enables fungibility.

I have never been an FBI agent. Never conducted an investigation. Never worn a wire or served a warrant or spent a winter in a field office where the heating runs four hours behind the interrogation schedule. I was a congressional staffer. Then a political appointee. Then a different kind of political appointee. Then the Director of the Federal Bureau of Investigation, which is also a kind of political appointee, except the furniture is nicer and the jet is mine. I run the building. I would like to tell you about the jet. It seats fourteen. It costs sixty million dollars. The interior was refurbished during the Bush administration and the procurement file describes the upholstery as "heritage cognac." I know this because I requested the file. Not for oversight purposes. I wanted to know the name of the color so I could describe it at dinner. Heritage cognac. It smells like a law firm that has never lost. I spend a lot of time in that smell now. I think it is the smell of having arrived somewhere that was never meant for you, and noticing that nobody has asked you to leave. Washington to Philadelphia is a hundred and forty miles. Amtrak runs it for forty-nine dollars. I flew the Gulfstream on May 10th because Alexis wanted to see George Strait. The suite was thirty-five thousand. Maybe fifty. I don't track numbers below six figures. The flight crew stayed on past eleven. Overtime. Security too. Someone will calculate the cost per mile of flying a sixty-million-dollar aircraft to cover a distance shorter than most Uber rides. That someone will not be me. I was in the suite. The suite didn't have a calculator. It had George Strait. The Bureau told reporters Alexis was "an invited guest of the performers." Representatives for George Strait and Chris Stapleton did not confirm this. They were never going to. But the FBI said it, and under my leadership, when the FBI says something, that is the evidentiary standard. I run the building. The building said it. It's true. Her protection detail is where the budget gets interesting. Twenty-four-seven coverage. SWAT-certified agents. Field officers drawn from multiple Bureau offices nationwide. Two armored SUVs at minimum. Hair appointments. Musical appearances. A blowout in Nashville required four agents in a parking lot for ninety minutes. The annual cost is roughly one million dollars before overtime, vehicle maintenance, and incidentals. The Bureau cites "hundreds of credible violent death threats" as justification. One person has been arrested. His name is Alden Welch Ruml, twenty-six, from Massachusetts. He sent emails. He faces five years. One emailer. One million dollars a year. Four SWAT agents per errand. A quarter of a million dollars in federal protection per verified threat. For context: the average FBI field office spends less annually on its entire confidential human source program. Informants embedded inside drug cartels, counterterrorism cells, organized crime families — the people whose testimony sends other people to prison — cost the Bureau less to keep alive than it costs me to ensure my girlfriend arrives at a salon escorted by men trained to rappel from helicopters. One of my officials suggested we conduct a formal threat assessment. A review. The kind of paperwork the Bureau has generated for seventy years whenever a protection decision requires justification. He had been with the Bureau for nineteen years. He asked in the hallway, in front of staff. I berated him. Not quietly. He stopped asking. He stopped asking. That phrase is going to recur in this document. I want you to notice it each time. Now. Alexis Wilkins is a country singer. I should mention this. She has released eight singles and one EP titled *Grit*. She has zero chart appearances. Her most-streamed song has five hundred and two thousand lifetime plays on Spotify. Taylor Swift accumulates that figure in approximately eleven minutes. A full-time country music journalist at the industry's leading independent publication stated that he had never heard of Alexis Wilkins before I started dating her. She has no active tour. She has never headlined a festival. Rolling Stone cited a forty-eight percent increase in her streaming numbers after we became public, which the same journalist clarified was "a forty-eight percent increase from virtually nothing." This is the person for whom the FBI fields a million-dollar annual protective detail. I want you to hold those two facts together — the streaming numbers and the security budget — and understand that they do not represent a miscalculation. They represent a value system. Her protection is not proportional to the threat. Her protection is proportional to her proximity to me. That is the formula. Proximity to the Director equals resources from the Bureau. I have applied it consistently. I have applied it to everyone. Which brings me to the people who are no longer here. I have removed over fifty career officials since January. Twelve have filed suit. Marshall Yates — my Congressional liaison, formerly the director of something called the Election Integrity Network, which existed to challenge the 2020 results — personally called field offices to compile names. Everyone who worked a case involving the President. Six thousand names were requested by the White House. The acting director before me was asked who he voted for. When he started supporting the President. Whether the FBI had tried to "put the president in jail." He was told the President hasn't forgotten. Three hundred counterterrorism and counterintelligence agents have been reassigned to immigration enforcement. The unit monitoring Iran — Iran, which operates proxy militias across four countries and maintains an active assassination program targeting American officials on American soil — was gutted. Six federal prosecutors in the Eastern District of Virginia have resigned or been pushed out rather than participate in the prosecution of the previous FBI Director, James Comey, whose crime was investigating the President and whose punishment is being investigated by the institution the President gave me as a gift. I am prosecuting the last Director for doing his job. I am doing this from a fifty-thousand-dollar suite while a sixty-million-dollar aircraft idles on the tarmac outside. Nobody in the building finds this ironic. The ones who would have found it ironic are gone. They stopped asking. My Deputy Director is Dan Bongino. He has never worked a federal case. His career before this was conservative talk radio. He receives the President's Daily Brief every morning — CIA product, NSA intercepts, the full intelligence take of the United States government — and he obtained his SCI clearance after I waived his polygraph. The FBI's own guidelines state that polygraphs are a "preliminary employment requirement." My lawyers reclassified him as a Schedule C political appointee. Experts said that's not how the statute works. The experts are career officials. Career officials are the previous administration's furniture. I am redecorating. Nikole Rucker is my personal assistant. She arrived at the Bureau on January 20th without a security clearance of any kind. She was physically escorted into the Director's suite because the door requires a clearance she did not possess. By February she was in London, seated across from a Western allied intelligence service, notebook open, pen moving. She used to work for Stephen Miller. The White House says she does not share operational details with him. I am told this is technically accurate in the way that most technically accurate statements are technically accurate. The polygraphs are still running. Just not for my people. We administer them now to career staff. The questions have changed. We ask whether they've criticized me. Whether they've spoken to a reporter. Whether they've expressed doubt about the direction of the Bureau. The machine measures stress. Under my leadership, stress has been reclassified as disloyalty. Disloyalty as a security risk. A security risk as grounds for termination. Fifty people have traveled this chain. Twelve are suing. The rest stopped asking. I run the building. In February a New York Times reporter named Elizabeth Williamson published details about the protective detail. I opened a preliminary inquiry. Federal stalking charges. We searched our databases for her information. The Department of Justice reviewed the file, found no legal basis, and terminated the inquiry. Called it retaliation. The Times' executive editor called it "a blatant violation of Elizabeth's First Amendment rights." I do not retaliate. I respond to threats. A journalist publishing accurate reporting about my personal use of public resources is, by my definition, a threat to operational security. My definitions are the ones that govern inside this building. I wrote the organizational chart. There is a framed copy on my wall. It has one name at the top. The Atlantic published a separate story. Excessive drinking. Frequent absences. Staff forcing entry into my home because I could not be reached. I filed a two-hundred-and-fifty-million-dollar defamation lawsuit. At my budget hearing, Senator Van Hollen cited the allegations under oath. I told him the only person slinging margaritas on the taxpayer dollar was him — in El Salvador, with a convicted gang-banging rapist. Fox News subsequently noted that public records do not support either characterization. But the line worked. That is the difference between evidence and performance. I have always understood which one this building rewards. In 2023, before any of this, I said the following on national television: "Chris Wray doesn't need a government-funded G5 jet to go to vacation. Maybe we ground that plane." I meant every word. We should have grounded his plane. So mine wouldn't invite the comparison. I sell merchandise. "Fight with Kash." T-shirts, hats, a children's book. The profits go to a foundation I started. The brand benefits from my position as Director of the Federal Bureau of Investigation. This is not a conflict of interest. A conflict requires two competing interests. I have one interest. It has never been healthier. I told the Senate that the FBI cannot meet its mission with a five-hundred-million-dollar cut. I requested twelve billion. Two billion more than last year. In the same period I spent a million on my girlfriend's security detail, fifty thousand on a concert suite, flew a sixty-million-dollar aircraft to cover a distance shorter than most commutes, waived background checks for three political appointees with no law enforcement experience, reassigned three hundred counterterrorism agents to check green cards, gutted the unit tracking Iran's assassination program, and opened a federal investigation into a newspaper reporter for the crime of publishing a newspaper. I told Hannity: "We are going to protect not only me and my loved ones but every American that is threatened." I meant the first seven words. The rest was institutional boilerplate. The kind of thing you say when the camera is on and the sentence needs to land somewhere that sounds like it includes other people. I run the building. Now I want to tell you about the water. The week before the concert I went to Pearl Harbor. The USS Arizona. A VIP snorkel. Nine hundred sailors and Marines are entombed in that hull. They have been there since 1941. The oil still leaks. It rises to the surface in small dark rainbows that break apart when you swim through them. The water was warm. Very clear. I could see the outline of the ship's superstructure below me, the geometry of a vessel that sank with its crew inside, and I remember the water temperature was perfect and the sun was on my back and my detail was on the shore and nobody in the water asked me to justify my presence above nine hundred dead. Recreational swimming at the Arizona is prohibited. The National Park Service said they were not involved. The Navy could not identify who authorized the outing. The logistics were coordinated by military email. A former government diver spoke to reporters anonymously. He said the access was unusual. He said it raised safety and security concerns. He spoke anonymously, the article noted, "for fear of retribution." A man who dives for the government is afraid to describe, on the record, how I swim. That is the climate. That is the building I run. A nineteen-year veteran stopped asking. Fifty career officials stopped working here. Three hundred counterterrorism agents stopped tracking the people who want to kill Americans. Six prosecutors stopped prosecuting. A government diver stopped talking. A reporter found her name in a database. And the oil keeps leaking from the Arizona, eighty-four years after the hull settled, surfacing in thin iridescent films that nobody is assigned to monitor because I reassigned them. I have never been an FBI agent. I have never conducted a federal investigation. I have never built a case or flipped a witness or spent a night in a surveillance van waiting for someone dangerous to make a mistake. But I have flown a sixty-million-dollar jet to a George Strait concert. I have watched the show from a suite that cost more than most Americans earn in a year. I have swum above nine hundred dead sailors in water so clear I could see their ship. And I have ensured, through the systematic removal of everyone who might object, that no one in the building will tell you any of this is wrong. The oil surfaces. It always surfaces. It has for eighty-four years. I run the building. The building doesn't ask questions anymore.

Former Intel CEO Pat Gelsinger with a16z's Raghu Raghuram and Guido Appenzeller on the next wave of semiconductor innovation and the physical constraints shaping the AI buildout: Drawing on his experience designing Intel's 386 and 486 processors, Pat explains how AI could transform chip design, but also why faster design alone won't solve the industry's biggest problems. They examine the bottlenecks in manufacturing, memory bandwidth, advanced packaging, and power, and why today's explosion of specialized AI chips may eventually consolidate around a smaller number of architectures. They also discuss the potential for new memory technologies, the shift from copper to optical networking, and why energy capacity could become a major constraint on AI growth. Finally, they revisit Pat's VMware years to ask what virtualization might look like when infrastructure is built for agents rather than humans. YouTube: 1:05 Hired by Intel at 18 3:50 Arguing with Stanford professors 7:25 Will AI design the next gen chips? 8:45 Designing faster than fabs can build 11:55 Why GPUs are stuck with bad memory 13:40 Why 100 AI chips will collapse to a few 19:15 Agents make new chips cheap to run 22:10 No new memory chip since the 90s 27:20 Why 16-layer chip stacks are a fantasy 32:15 The physics that keeps memory next to compute 34:55 "I declared the death of copper" 38:05 AI traffic breaks the internet's design 42:40 "Energy capacity equals economic capacity" 46:25 The 130-year-old power war AI reopened 48:50 VMware for agents @PGelsinger @RaghuRaghuram @appenz

one of the most unintuitive problems to solve for in ml inference is the routing problem imo. and i claim that this, and this alone, is responsible for the birth of every flavor of sparse attention (SWA, NSA, DSA, etc). to set up why routing birthed sparse attention: you have multiple nodes, and you have multiple requests. 1) can do pure load-balancing: request 1 goes to node 1 request 2 to node 2 (node 1 is busy) request 3 to node 3.... this would be perfect if the model did not depend on the kv cache. but it does, and so this is a terrible system, since you do prefill over the same tokens many times. on the other extreme: 2) can do pure cache-aware routing. request 1 goes to node 1 ( kv cache stored on node 1) request 2 goes to node 1 (the kv cache is on node 1) request 3 goes to ... node 1 (the kv cache ...) this would be perfect if the system only depended on the kv cache. but it does not. decode time makes the requests queue, so this is also a terrible system. and so it seems that the two pull in opposite directions (at least if we run both the cache aware and load based routing to their extreme case) this is analogous to a somewhat dysfunctional engineering team. if ticket i+1 comes up, do you give it to the person who has context on ticket i, or do you route it to someone who has to onboard? the former is probably working on something else, but could finish it quickly. the latter can start on it immediately, but will take a while. you have to do a balance of the 2. and so the solution becomes a calculation that the router has to do: what is the a) estimated queue time b) number of uncached tokens c) speed of prefill (processing said uncached tokens) and it assigns the total=a+b/c to each node, picking the node with the least total time at a per request level. but, imo, this is a patch that exists because the kv cache stays on the node it was computed on, because we don't have the ability to keep one shared kv cache / one shared storage system for every node within a cluster... but i'm obviously wrong. mooncake (and others) exists. maintaining a shared kv pool is a decade old idea... but then: 1) you're still limited by the fact that you can't have a remote / distributed kv pool. you need every single block in the kv to be present on the node you're doing decode on (due to HBM speed and decode being mem bound), so you must pay the price of doing the transfer. 2) the global pool can only try to transfer some of the kv to a node, but cannot guarantee, because of evictions, and long transfers at scale. even if transfers were guaranteed, even if evictions never happened, you still need (b)/(c) for the 'redo prefill' vs 'wait for kv transfer' arbitrage. so then: - what is the point of maintaining global kv pool, if only to track where distributed kv blocks are and do said arbitrage, if it cannot be used to directly draw kv blocks from? - can this only be solved when network bandwidth equals that of HBM (ie unsolvable problem)? and, on the premise that the answer to both points above is yes, the solution becomes to need the least amount of kv blocks per request, and this is why every open source lab is racing to show off a new and shiny sparse attention flavor.

My coworker Morimoto's father is 86. A rice farmer in a small village in Niigata. For 62 years, he has been doing things nobody in that village understands. Every Saturday morning, he bakes an apple pie. And eats it for breakfast. With black coffee. At 5 a.m. His wife fought him on this until 1971. Then she gave up and started eating it too. He says "You bet" to everyone. It's the only English he uses. He said it to the mayor. He said it to the Buddhist priest at his own mother's funeral. The priest has learned to say "You bet" back. When a child is introduced to him, he stands up, puts out his hand, and shakes it like they're a grown man closing a deal. "Pleased to meet you." Village kids have been lining up for this since 1965. Some of them are grandparents now. They still line up. When a woman walks into a room, he stands. Every time. Every woman. The old ladies in the village giggled about it for sixty years. Now they wait in the doorway until he gets up. If something is broken at a neighbor's house, it gets fixed. A gate. A gutter. A tractor. He doesn't ask. He doesn't mention it. He's already walking home. Every Sunday at 7 p.m., he calls each of his three children. In English. He says three words. "Just checkin' in." Then he hangs up. Morimoto is 58. He has received that phone call every Sunday for forty years. He has never once been able to say anything back in time. And every night at supper, in a Buddhist farmhouse in rural Japan, the old man bows his head and says grace. In English. He closes his eyes. He thanks God for the rice and the rain and the family at the table. And he always ends the same way. "And thank you for Leonard and Ruth. Amen." Then he claps his hands twice. The way we do at a Shinto shrine. Nobody in the village knows who Leonard and Ruth are. They've been hearing their names every night since 1965. In 1964, Morimoto's father was 24, and the poorest young man in his village. A program was sending young Japanese farmers to live on American farms for a year. He went with one suitcase and almost no English. It was only nineteen years after the war. His own uncle had died in it. He'd grown up hearing what Americans were. He spent the whole flight across the Pacific sure he was going to be hated. He had decided he would bear it quietly. Work hard. Bow. Disappear. The farm was in Iowa. Corn and hogs. Flat to the edge of the world. Leonard met him at the bus station. Six-foot-three, forty years old, hands like shovels. A Marine Corps tattoo on his forearm, faded green. He had fought on Okinawa. He looked at the frightened young man for a long time. Then he put out his hand and shook it like he was a grown man. Like they were equals. "Pleased to meet you, son." That night, at the first supper, Leonard bowed his head and said grace. Tsutomu didn't understand most of it. But he understood the last line, because Ruth leaned over and said it again for him, slowly. "And thank you, Lord, for bringing us Tsutomu." He cried into his mashed potatoes. Nobody said a word about it. Ruth just passed him more gravy. For a year, he was their son. Ruth taught him pie. Leonard taught him to drive a tractor, then a truck, then to back a trailer, which took four months and a lot of yelling, which Leonard later apologized for by fixing Tsutomu's bicycle without being asked. Leonard drove three hours to Des Moines every month to buy a bag of rice so the boy wouldn't get homesick. Every Sunday night, Leonard called his own grown daughter in Omaha. "Just checkin' in." Then hung up. On his last day, at the bus station, Leonard shook his hand like a grown man. And then, for the first and only time, he hugged him. Hard. The way a man hugs when he's not used to it. "You come home anytime, son." Tsutomu never could. He went back to Japan and became a farmer and a father and a grandfather. He wrote every Christmas. Leonard wrote back every Christmas until he died in 1993. Ruth wrote until 2001. And for sixty-two years, the old man in Niigata kept doing exactly what Leonard did. The pie. The handshake. The standing up. The fixing. The phone call. The grace. All of it. Like a man carrying a lantern across an ocean, very carefully, so it wouldn't go out. Last spring, Morimoto found Leonard's granddaughter on Facebook. Her name is Jenny. She's 46. She still lives in that county. He sent her a video. His father at the supper table, eyes closed, saying grace. In English. "And thank you for Leonard and Ruth. Amen." Clap. Clap. Jenny answered two days later. She said she'd watched it thirty times. She said she had to sit down on her kitchen floor. Then she sent a photo. The refrigerator in the old farmhouse. Jenny's mother lives there now. Nobody has ever changed that refrigerator door. Two photos are taped to it. Yellow. Curling at the edges. Held on with tape that has been replaced many times. One is a young Japanese man in overalls, squinting into the sun, standing next to a tractor. 1964. The other is a young Marine in uniform. Nineteen years old. Smiling. His name was Eddie. He was Leonard's little brother. He was killed on Okinawa in 1945. Leonard was there. He came home without him. And then Jenny told him the part that nobody in Japan had ever known. In 1963, when the letter came saying a young Japanese man would be living with them for a year, Leonard didn't say anything. He went out to the barn after supper. He didn't come back in. Ruth found him there at dawn. Sitting on a hay bale. Holding Eddie's picture. She didn't know what he would say. She was afraid of what he would say. He stood up. He put the picture in his shirt pocket. And he said: "Well. He's somebody's son too." Then he walked into the house, got out a can of paint, and started painting the spare room. It was Eddie's room. That's where Tsutomu slept. For a whole year. In the bed of the brother Leonard lost. He never knew. Jenny said her grandpa never once talked about the war. Not to anyone. Not ever. But he talked about Tsutomu all the time. Morimoto drove to Niigata that weekend and read the message to his father at the kitchen table. Slowly. Translating every line. The old man didn't say anything for a long time. Then he got up, walked out the back door, and stood in the rice field until it was dark. At supper that night, his whole family was at the table. His children. His grandchildren. His great-grandchildren. Fourteen people. He bowed his head. He closed his eyes. He thanked God for the rice and the rain and the family at the table. "And thank you for Leonard and Ruth." Then he stopped. The whole table waited. When he spoke again, his voice was very quiet. In the English he had learned in Iowa, sixty-two years ago. "And thank you for Eddie. I slept in his room. I tried to take good care of it." Amen. Clap. Clap.

I have a dream. That one day $ANSEM will rise from the trenches and live out the true meaning of its creed: “We hold these truths to be self-evident, that all bags are created equal.” I have a dream that one day, on the green candles of Solana, the sons of former jeets and the sons of former diamond hands will sit down together at the table of generational wealth. I have a dream that one day, even CT, a place sweltering with the heat of FUD, rugs, and endless PvP, will be transformed into an oasis of green candles, 100x returns, and financial freedom. I have a dream that one day, my fellow trench warriors will live in a world where their bags will not be judged by the color of their PnL, but by the strength of their conviction. I have a dream that one day the cabals will be defeated, the jeets will be forgiven, and every last holder will witness the promised land of price discovery. I have a dream that one day, every red candle shall be reversed, every resistance shall be broken, and the glory of the god candle shall be revealed for all of CT to see. I have a dream today. $ANSEM 🐂🀄️