Discussion | Link

Sean Michael Kerner / VentureBeat: Moonshot AI releases Kimi K2.7-Code, claiming 30% lower reasoning token usage compared to K2.6, available under a modified MIT license  —  Moonshot AI released Kimi K2.7-Code this week, an open-source update to its K2 coding model family, claiming leaner reasoning and double-digit performance gains.

Vicky Ge Huang / Wall Street Journal: Crypto-data provider Blockworks says it has acquired its rival Messari; source: Blockworks paid more than $10M for Messari, which was valued at $300M in 2022  —  Buyer pays more than $10 million for company valued at around $300 million in 2022  —  The crypto-data provider Blockworks …

<p><strong><a href="https://www.anthropic.com/news/fable-mythos-access">Statement on the US government directive to suspend access to Fable 5 and Mythos 5</a></strong></p> Well this is <em>nuts</em>:</p> <blockquote> <p>The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for <strong>all</strong> our customers to ensure compliance. <strong>Access to all other Anthropic models</strong> <strong>will not be affected.</strong></p> <p>We received the directive from the government today at 5:21pm (ET). The letter did not provide specific details of its national security concern. Our understanding is that the government believes it has become aware of a method of bypassing, or "jailbreaking" Fable 5. We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass. [...]</p> <p>To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws. Our understanding is that one potential jailbreak was shared with the government. We have reviewed the report and validated that the level of capability displayed there is widely available from other models (including OpenAI's <a href="https://deploymentsafety.openai.com/gpt-5-5/tacit-knowledge-and-troubleshooting">GPT-5.5</a>), and is used every day by the defenders who keep systems safe. We will share more details over the next 24 hours.</p> </blockquote> <p>I still have access to Fable via <a href="https://claude.ai/">claude.ai</a> and Claude Code now, at 9:01pm ET.</p> <p><strong>Update</strong>: I ran <a href="https://gist.github.com/simonw/5894cfafc64a2b8aafbe834bc9c950b9">this script</a> against the Anthropic API to spot when <code>claude-fable-5</code> would stop working. My access was cut off at 6:59pm Pacific (9:59pm ET):</p> <pre>[2026-06-12T18:56:50-07:00] attempt 35: running uv run llm -m claude-fable-5 hi [2026-06-12T18:56:55-07:00] success: Hi there! How can I help you today? [2026-06-12T18:57:55-07:00] attempt 36: running uv run llm -m claude-fable-5 hi [2026-06-12T18:57:59-07:00] success: Hi! How can I help you today? [2026-06-12T18:58:59-07:00] attempt 37: running uv run llm -m claude-fable-5 hi [2026-06-12T18:59:00-07:00] FAILED after attempt 37 with exit code 1 stderr: Error: Error code: 404 - {'type': 'error', 'error': {'type': 'not_found_error', 'message': 'Claude Fable 5 is not available. Please use Opus 4.8. Learn more: https://www.anthropic.com/news/fable-mythos-access'}, 'request_id': 'req_011CbzRyirV7KZLHYYdBM9od'}</pre> <p><small></small>Via <a href="https://twitter.com/AnthropicAI/status/2065597531644743999">@AnthropicAI</a></small></p> <p>Tags: <a href="https://simonwillison.net/tags/jailbreaking">jailbreaking</a>, <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/anthropic">anthropic</a>, <a href="https://simonwillison.net/tags/claude">claude</a>, <a href="https://simonwillison.net/tags/ai-ethics">ai-ethics</a>, <a href="https://simonwillison.net/tags/claude-mythos">claude-mythos</a></p>

Anthropic: Anthropic says it is disabling Fable 5 and Mythos 5 for all customers after the US government issued an export control order, citing national security concerns  —  The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 …

Brickster Voices is a series that spotlights the people who make our work possible. Through personal career journeys...

Axios: Source: the US Commerce Department decided to block foreign persons from accessing Fable 5 and Mythos 5 after a company claimed it was able to jailbreak Mythos  —  The Trump administration is blocking foreign governments, companies and individuals from accessing Anthropic's most advanced AI models.

<p><strong><a href="https://tools.simonwillison.net/openai-webrtc">OpenAI WebRTC Audio Session, now with document context</a></strong></p> I built the first version of this tool <a href="https://simonwillison.net/2024/Dec/17/openai-webrtc/">in December 2024</a> to try out the then-new OpenAI WebRTC API for interacting with their realtime audio models.</p> <p>Last month OpenAI <a href="https://openai.com/index/advancing-voice-intelligence-with-new-models-in-the-api/">introduced a brand new model</a> to that API called <a href="https://developers.openai.com/api/docs/models/gpt-realtime-2">GPT‑Realtime‑2</a>, which they promoted as "our first voice model with GPT‑5‑class reasoning" - with a Sep 30, 2024 knowledge cut-off.</p> <p>I've been waiting for that model to show up in the ChatGPT iPhone app but it still hasn't, so I revisited my old playground.</p> <p>You can now pick the better model, and you can also paste in a big chunk of document context so you can have as audio conversation in your browser about whatever information you think would be useful to explore in a conversational way.</p> <p><img src="https://static.simonwillison.net/static/2026/openai-webrtc-document-context.jpg" alt="Screenshot of a web interface titled &quot;OpenAI WebRTC Audio Session&quot; with a gray status dot. Form fields: &quot;OpenAI API Token&quot; showing a masked password of dots, &quot;Voice&quot; dropdown set to &quot;Coral&quot;, &quot;Model&quot; dropdown set to &quot;gpt-realtime-2&quot;. A collapsible section labeled &quot;▼ Document context (optional — paste text to talk about)&quot; with bold instruction &quot;Paste a document here before starting the session and the model will be able to discuss it with you&quot; above a textarea containing a pasted Markdown document about whether DuckDB can run untrusted SQL as safely as Datasette runs SQLite. Below are a blue &quot;Start Session&quot; button and a gray disabled &quot;Mute Mic&quot; button, then a green success message &quot;Session established successfully!&quot; At the bottom, a dark panel headed &quot;Last transcript&quot; reads: &quot;DuckDB can be made about as safe as SQLite for running untrusted SELECT queries, but only if you lock it down properly. Using read only true by itself is not enough, because SQL can still&quot; (text cut off)." class="blogmark-image" style="max-width: 80%"> <p>Tags: <a href="https://simonwillison.net/tags/audio">audio</a>, <a href="https://simonwillison.net/tags/tools">tools</a>, <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/openai">openai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/multi-modal-output">multi-modal-output</a>, <a href="https://simonwillison.net/tags/webrtc">webrtc</a></p>

“I’m not sure that this company supports a hackathon culture anymore,” one employee posted in a forum open to the entire staff.

Wall Street Journal: OpenAI was subpoenaed on Friday by a coalition of state AGs for documents covering a wide range of its activities and impact on users  —  Company subpoenaed for documents covering wide range of its activities and impact on users  —  A coalition of state attorneys general has opened …

Wired: Sources detail frustration inside Meta's Applied AI team, formed in March to support Superintelligence Labs, over menial projects, “soul-crushing” work, more  —  Executives and employees alike are struggling with Meta's chaotic AI strategy, according to sources and internal discussions reviewed by WIRED.

As of today, SpaceX is owned by investors who will want to see it make money.

Julius Mondragon, who goes by "Computah" on TikTok, is using Meta's Ray-Bans in the best way yet. Katie Notopoulos/Business Insider Julius Mondragon is "Computah." He goes around "programming" people on the streets of Chicago. His TikTok videos have gone viral. They've made him a local hero. It's one of the few good use cases I've seen for Meta's Ray-Bans. On a chilly winter day, Julius Mondragon stepped into a half-empty car on the Metra train line in Chicago. "Computah, activate hyperspeed typing," he commanded a random woman typing away on her laptop. "Computah, make these two guys friends forever," he said to two men sitting together, as they looked up, smiling. "Computah, make sure this man feels confident and sexy at all times of the day," he instructed. "Computah, make sure this woman has a wonderful day." Before opening the door between cars, heading off in his Meta Ray-Ban glasses to find the next group of unsuspecting commuters, he announces: "You've been programmed!" Mondragon has been "programming" people all over Chicago — and beyond. The 26-year-old has gone viral on TikTok while summoning his weirdly accented "Computah" character, wearing Meta's Ray-Ban glasses and recording people's reactions to his bot-like commands. I think it's the best use of Meta's Ray-Bans that I've seen, so I asked to meet up when he was in New York this week so I could see the "Computah guy" in action. Meeting 'Computah' in real life I met up with Mondragon in a Harlem café near the hotel where he was staying. Unlike the bizarre persona in his videos, he's genial and soft-spoken. (The voice he does in his videos is a fake accent based on an imitation of a character in "The Sopranos" dreamed up by him and his roommates.) This is where the Meta Ray-Bans come in: Mondragon got a pair this past Christmas. He started posting videos recorded by the glasses to TikTok in January. The "Computah" idea initially came from an Eric André sketch where André, wearing a black leotard with electronic parts glued to it, pretends to be Google as he approaches strangers on the street. "When I got the Metas, I was coming up with ideas," Mondragon told me. "What if I'm Meta and I'm pretending I'm just an AI and I can transform anything in my surroundings?" That's where "Computah" came from. Mondragon grew up in the Chicago suburbs and currently lives on the South Side of Chicago. For the last few years, he's been a music producer — something he still pursues — while working a day job in retail. He has recently been able to quit his day job at Foot Locker thanks to the money he's earned through social media. He told me that Cameo, where he charges $40 to send a personalized message to fans, has also been particularly helpful in getting him to a financial position where he can pursue "Computah" full-time. Along the way, he's become something of a Chicago star. He was profiled in the Chicago Tribune and appeared on the local TV news. In his videos, it's clear people are recognizing him, leading to funny interactions where they mimic his catchphrases and tactics, like saying "Firewall up!" to block his "programming." The Meta Ray-Bans are something of a reverse Clark Kent for Mondragon — people only recognize him when he puts on the glasses. Meta Ray-Ban glasses make the surveillance state funny These smart glasses are not without controversy. From their launch in 2021, there has always been some concern about people using them as "creep glasses" to record people surreptitiously. Since then, Meta's glasses have sold decently. In 2025, Meta said it sold more than 7 million pairs — triple the amount it sold in 2024. Presumably, people use them for private photos and videos, but some of the footage that gets posted onto social media tends to fall into two buckets: trolls who make videos harassing service workers, and men making videos of themselves trying to pick up women on the street. Neither is a particularly positive outcome for the technology. "My feed is mostly Meta [videos]. It's just mainly pickup artists and people trying to ragebait people," Mondragon said. "I knew when I started, I didn't want to do that. Obviously, I'm looking to get a reaction, but I'm not looking to mess up this person's day. I didn't want it to be like, this person leaves this interaction feeling shitty about themselves or anything." When asked about "Computah's" popularity, Meta spokesperson Andy Stone told Business Insider: "Local hero wears Ray-Ban Metas is a great story!" And "Computah" is funny. It's hard to explain why something is funny other than just pointing out, "This is funny." I mean, you all get the joke. But I'll try to explain what I think makes it work so well: Doesn't it feel like we're all being "programmed" somehow by some unseen computer? That our worldview and the trajectory of our days, and even our bodies, are being shaped by some unfeeling algorithm? Sure it does! Mordragon sees a deeper meaning to it beyond just being funny. "It's like a pattern interruption," he said. "That's really what I'm trying to go for." Mondragon (right) "programs" a fan who recognized him in Washington Square Park in New York City. Katie Notopoulos/Business Insider 'Computah' has one weird catch phrase that's caught on A particular "Computah" phrase has become a meme in its own right. One of Mondragon's catch phrases has been a lighthearted command: "Computah, make these guys super gay and horny!" (He made sure to tell me he means this in the spirit of fun, not as something antigay.) Another TikToker made a catchy song using the lyric over and over. Mondragon told me he wasn't really sure why that particular phrase has caught on so much, or why girls seem to have latched onto it. Personally, I think it hits something in the gender and sexuality zeitgeist that simply just works in 2026. View this post on Instagram A post shared by computa (@computahh) On the day we met this week, the comedian and YouTuber Druski announced a rally for the New York Knicks in Washington Square Park. Mondragon and I took the subway downtown together from Harlem. Just before our final stop on the train, he donned his Meta Ray-Bans and started to program the people in the car. As you might expect from typical New Yorkers, they didn't react much to the man speaking loudly and saying strange things on the subway. When we reached the rally, where a few hundred young men had gathered to chant various permutations of "Knicks in four," we hung on the outskirts of the crowd. Suddenly, someone recognized him: "Hey, it's Computah!" and excitedly shook his hand. A young couple visiting from Florida, Amanda Rivera and her boyfriend, Diego Casanova, approached "Computah" and asked him to program them. Casanova told me that he was inspired by Mondragon to buy his own pair of Meta glasses. (He was wearing the frames with clear lenses while I chatted to him. I didn't even realize he was wearing them until he mentioned it. His recording light was on, but he didn't seem to realize he'd been recording our interaction.) Mondragon started programming the Florida visitors, and heads turned. More cheers erupted as people recognized him, and a crowd formed. The people who showed up to a spontaneous gathering for a social media comedian were the perfect chronically online cohort who had also all seen Mondragon's antics on TikTok and Reels. While he might slip by on city streets unnoticed, in this crowd, he was a star. Read the original article on Business Insider

Paramount Skydance CEO David Ellison has DOJ clearance to buy Warner Bros. Discovery. Angela Weiss/AFP via Getty Images; Dania Maxwell / Los Angeles Times via Getty Images Paramount Skydance got the US Department of Justice's approval to acquire Warner Bros. Discovery. "The transaction is not likely to result in harm to competition or American consumers," the DOJ said. Buying WBD's studio and streaming assets would make Paramount a Hollywood superpower. David Ellison's Paramount Skydance just got the official green light to buy Warner Bros. Discovery from President Donald Trump's Department of Justice. The US government signed off on Paramount's $111 billion mega-deal on Friday, removing a major obstacle for Ellison's media company as it tries to build a Hollywood superpower. However, Paramount could still face regulatory challenges abroad or lawsuits by US states. "The transaction is not likely to result in harm to competition or American consumers" in streaming, traditional TV, or film production and distribution, the US DOJ said in a statement. The DOJ also said that "substantial evidence does not suggest a likelihood of reduction in output" in creative output. "We are grateful for the Department of Justice's thorough review of this transaction, as well as the work of the other agencies that have completed their reviews and provided clearance to date," a Paramount spokesperson said in a statement. Paramount has said it's aiming to acquire WBD by the end of September. It's agreed to pay WBD shareholders a so-called "ticking fee" of about $7 million per day that the deal isn't closed, starting September 30. WBD had originally agreed to sell its studio and streaming assets, including the Warner Bros. studio and HBO Max, to Netflix for $27.75 per share. Paramount responded by offering $30 per share for the entire company, including its TV assets like CNN, HGTV, and TruTV. Both Paramount and Netflix argued that they had the more favorable regulatory path and were offering more value to WBD's investors. WBD's board decided in February that the Paramount offer was better than Netflix's. Read the original article on Business Insider

Executives and employees alike are struggling with Meta’s chaotic AI strategy, according to sources and internal discussions reviewed by WIRED.

Discussion | Link

AI agents have fundamentally changed the complexity of inference workloads. Until now, the industry has struggled to define a standard for measuring how... AI agents have fundamentally changed the complexity of inference workloads. Until now, the industry has struggled to define a standard for measuring how inference systems perform under these conditions. Artificial Analysis AgentPerf (AA-AgentPerf) offers the industry’s first multi-vendor open benchmarks profiling trajectories that are representative of real-world AI agent coding tasks. Source

Bloomberg/Getty Images Zohran Mamdani called out Elon Musk's trillionaire status on Friday. Mamdani said it was the trillionth reason to tax the rich. Musk became the world's first trillionaire after the SpaceX IPO. New York City Mayor Zohran Mamdani used SpaceX's historic IPO — and founder Elon Musk's newly minted status as a trillionaire — to push policy. "Reason #1,000,000,000,000 why we should tax the rich," Mamdani wrote on X in response to Musk becoming the world's first trillionaire on Friday. Musk reached the financial milestone after SpaceX stock popped upon hitting the market on Friday, trading at $150 per share, above the company's initial offering of $135. Already the world's richest man, the windfall put Musk leagues ahead of the next richest person, Google's Larry Page, who now has a quarter of Musk's wealth. Mamdani's response was in line with his politics. The progressive New York City mayor campaigned on taxing the rich, and the city's new pied-à-terre tax on multimillion-dollar second homes sparked backlash from some rich residents. Musk has spoken out against Mamdani and backed his challenger, Gov. Andrew Cuomo, in the New York mayoral race last year. Mamdani has said that he and Musk share one goal: focusing on efficiency and eliminating waste. Mamdani last month announced he was launching a Commission on Government Efficiency, essentially New York City's answer to the Musk-led DOGE office in the White House. Read the original article on Business Insider

Financial Times: KPMG retracts a report on AI's benefits after it has been found to exaggerate AI adoption with case studies that appear to have been based on AI hallucinations  —  Bogus case studies on UBS and transit systems exaggerated adoption of the technology.  A KPMG report on how AI is being used …

Elon Musk's rocket company went public in a blockbuster IPO, skyrocketing the CEO's net worth above $1 trillion. How will the SpaceX IPO impact markets? Where does the company go from here? Business Insider's Dan DeFrancesco and Joe Ciolli have the answers to these questions and more in a live Q&A. Read the original article on Business Insider