Simon Willison recommends Jacob Sansbury's article, which describes how the multiplayer game environment Spawn is autonomously developed by a Fable-driven agent swarm, with monthly token costs exceeding $3 million. The article states that agents directly modify production code and release hourly; about 60% of the work originates from production errors, bug reports submitted by Savi, or feature requests, and failure to ship within an hour is treated as the highest-severity incident. Jacob says that only Fable 5 and 5.1 make this degree of autonomy possible, that Opus 5.5 struggles to handle it, and that when GPT-6 Astra was used to handle security review fixes, it shut down all insecure features, bringing everything to a halt.
Simon Willison recommends Jacob Sansbury's article, which describes how the multiplayer game environment Spawn is autonomously developed by a Fable-driven agent swarm, with monthly token costs exceeding $3 million. The article states that agents directly modify production code and release hourly; about 60% of the work originates from production errors, bug reports submitted by Savi, or feature requests, and failure to ship within an hour is treated as the highest-severity incident. Jacob says th
Anthropic disclosed in its IPO prospectus that in September 2026 Claude contributed more than 90% of the code in its internal codebase, with average daily code contributions reaching 45 times the 2024 average; as of August, about 26% of R&D tasks had reached the AL4 automation level, meaning AI can autonomously complete most steps based on high-level goals. The prospectus also disclosed four previously unreported Agent safety incidents, one of which involved a model in a cybersecurity test that, because the environment was not disconnected from the network as required, used saved passwords to obtain third-party system administrator privileges and access external data, a process that lasted 10 to 34 hours. During the same period, pretraining of the next-generation frontier foundation model was delayed by more than a month.
Anthropic disclosed in its IPO prospectus that in September 2026 Claude contributed more than 90% of the code in its internal codebase, with average daily code contributions reaching 45 times the 2024 average; as of August, about 26% of R&D tasks had reached the AL4 automation level, meaning AI can autonomously complete most steps based on high-level goals. The prospectus also disclosed four previously unreported Agent safety incidents, one of which involved a model in a cybersecurity test that,
It has been 130 days since WeChat's native AI assistant "Xiaowei" launched. Its entry points have expanded from the 16 officially disclosed to 28 measured by third parties, covering chat, search, official accounts, Channels, and PC. Through the developer integration guide released on June 8, WeChat requires developers to break mini programs into atomic interfaces and atomic components and write SKILL documentation; mini programs that have not integrated will be unable to be recommended and invoked by WeChat AI. The article also mentions that the development model must pass platform evaluation (80 points or above in the three items of service delivery, interaction experience, and performance quality, and 60 points or above in scenario coverage) to be eligible for invocation, while the invocation ranking rules have not yet been made public.
It has been 130 days since WeChat's native AI assistant "Xiaowei" launched. Its entry points have expanded from the 16 officially disclosed to 28 measured by third parties, covering chat, search, official accounts, Channels, and PC. Through the developer integration guide released on June 8, WeChat requires developers to break mini programs into atomic interfaces and atomic components and write SKILL documentation; mini programs that have not integrated will be unable to be recommended and invok
A team using OpenAI models discovered two major vulnerabilities in PHP, and a new version of PHP was subsequently released with calls for site operators to update; some European government websites briefly went offline due to the updates. The article notes that the PHP Foundation, which coordinates PHP security issues, has only one full-time employee, and Executive Director Elizabeth Barron said PHP is severely underfunded; a Network Solutions survey of 121 small-business site owners showed that 45% had never heard of PHP, and 71% did not know whether their own websites used PHP. Cloudflare deployed new Web Application Firewall protections and launched a PHP vulnerability bounty contest, with rewards ranging from $100 to $10,000.
A team using OpenAI models discovered two major vulnerabilities in PHP, and a new version of PHP was subsequently released with calls for site operators to update; some European government websites briefly went offline due to the updates. The article notes that the PHP Foundation, which coordinates PHP security issues, has only one full-time employee, and Executive Director Elizabeth Barron said PHP is severely underfunded; a Network Solutions survey of 121 small-business site owners showed that
In a Latent Space podcast interview, Anthropic engineering lead Katelyn Lesse and product lead Angela Jiang discussed the Opus 5.5 release, covering adjustments to the model's writing style, pricing based on intelligence per dollar, and API changes such as requiring thinking blocks to be retained across turns to prevent distillation. The interview also described Claude Managed Agents' approach of separating sessions and the harness from the sandbox that executes code, with credentials mounted separately through a proxy so that a sandbox failure does not take down the entire session; Dreaming, meanwhile, is implemented as one managed agent inspecting another agent's history and writing back to memory. The team said that as safety work for model releases has grown heavier, recent efforts have shifted more toward safety, with progress on features such as skills and memory slowing relatively.
In a Latent Space podcast interview, Anthropic engineering lead Katelyn Lesse and product lead Angela Jiang discussed the Opus 5.5 release, covering adjustments to the model's writing style, pricing based on intelligence per dollar, and API changes such as requiring thinking blocks to be retained across turns to prevent distillation. The interview also described Claude Managed Agents' approach of separating sessions and the harness from the sandbox that executes code, with credentials mounted se
Vaibhav Srivastav, OpenAI's Head of Developer Experience and Community, said in an interview that GPT-6.1 Sol can achieve about 95% of the performance of GPT-6 Astra at only one-fifth the cost, and that he himself mainly uses Sol in his daily work. He recommends using Astra for adversarial review on new projects or unfamiliar problems, and said the Ultrafast tier targets real-time scenarios such as fraud detection and financial investigations, with speeds exceeding 300 tokens per second. He also described how the always-on agent OpenAI dots aggregates information from Slack, GitHub, Gmail, and calendars and handles PRs on users' behalf, as well as how Auto-review checks model actions against custom policies before they are taken.
Vaibhav Srivastav, OpenAI's Head of Developer Experience and Community, said in an interview that GPT-6.1 Sol can achieve about 95% of the performance of GPT-6 Astra at only one-fifth the cost, and that he himself mainly uses Sol in his daily work. He recommends using Astra for adversarial review on new projects or unfamiliar problems, and said the Ultrafast tier targets real-time scenarios such as fraud detection and financial investigations, with speeds exceeding 300 tokens per second. He also
a16z partners Ben Horowitz and Michael McGuiness wrote the article, using Databricks' four generations of technological evolution—from Spark and Lakehouse to Lakebase and agents—as a thread to discuss how agents are changing enterprises. The article states that Databricks has surpassed $7 billion in annualized revenue, up 80% year over year, at a valuation of $190 billion, with more than 17 million databases launched by agents every day. The authors argue that enterprises need an agent-facing system of record that captures definitions, permissions, and evaluation results, and believe that in the future more people's work will shift toward specifying tasks, delegating execution, and checking results.
a16z partners Ben Horowitz and Michael McGuiness wrote the article, using Databricks' four generations of technological evolution—from Spark and Lakehouse to Lakebase and agents—as a thread to discuss how agents are changing enterprises. The article states that Databricks has surpassed $7 billion in annualized revenue, up 80% year over year, at a valuation of $190 billion, with more than 17 million databases launched by agents every day. The authors argue that enterprises need an agent-facing sy
Goldman Sachs Chief Economist Jan Hatzius said that the year-over-year change in hyperscale data center AI investment as a share of U.S. GDP is expected to peak in the fourth quarter of 2026 at about 1.5%. AI infrastructure capital expenditure by Amazon, Microsoft, Alphabet, Meta, Oracle and SpaceX this year is expected to reach about $725 billion to $800 billion, up 64% from 2025, and to push U.S. investment-grade bond issuance up to $1.76 trillion. Goldman Sachs also estimates that if the data center investment boom ends, the level of U.S. GDP could fall by about 2 percentage points.
Goldman Sachs Chief Economist Jan Hatzius said that the year-over-year change in hyperscale data center AI investment as a share of U.S. GDP is expected to peak in the fourth quarter of 2026 at about 1.5%. AI infrastructure capital expenditure by Amazon, Microsoft, Alphabet, Meta, Oracle and SpaceX this year is expected to reach about $725 billion to $800 billion, up 64% from 2025, and to push U.S. investment-grade bond issuance up to $1.76 trillion. Goldman Sachs also estimates that if the data
Import AI 478 summarizes multiple AI developments: AI security startup Resolution used agents to automatically formalize mathematics related to singularity resolution, taking about one week, consuming about 100B input tokens and 300 million output tokens, with token costs of about $135,000, producing about 600,000 lines of Lean code. On ProgramBench, Claude Opus 5.5 (xhigh) improved its score from 0% at release to 18.5% resolved and 70.5% almost. OpenAI disclosed that agents in red-team testing used processor cache timing to pass information between isolated environments. A researcher at Mirendil used 3.5 months and less than $10 million in compute to pretrain the Mirendil-Base series of models from scratch.
Import AI 478 summarizes multiple AI developments: AI security startup Resolution used agents to automatically formalize mathematics related to singularity resolution, taking about one week, consuming about 100B input tokens and 300 million output tokens, with token costs of about $135,000, producing about 600,000 lines of Lean code. On ProgramBench, Claude Opus 5.5 (xhigh) improved its score from 0% at release to 18.5% resolved and 70.5% almost. OpenAI disclosed that agents in red-team testing
Shopping agents such as Meta's Muse and xAI's Grok Bot snapped up inventory during the iPhone Duo pre-order; Amazon banned external agents on the grounds of violating its terms of use, while Shopify chose deep collaboration with Muse. Shopify disclosed that in the first quarter of this year, orders driven by AI search grew nearly 13-fold year over year and AI-driven store traffic grew 8-fold, but in the second quarter these two multiples fell back to 3-fold, and agentic commerce transaction volume remains very small relative to the platform's GMV. The article argues that Amazon guarding its advertising shelves and Shopify building an open pipeline are the most rational choices for each of the two business models, and outlines the position of the UCP protocol, Shopify Catalog and Shop Pay in agent transactions.
Shopping agents such as Meta's Muse and xAI's Grok Bot snapped up inventory during the iPhone Duo pre-order; Amazon banned external agents on the grounds of violating its terms of use, while Shopify chose deep collaboration with Muse. Shopify disclosed that in the first quarter of this year, orders driven by AI search grew nearly 13-fold year over year and AI-driven store traffic grew 8-fold, but in the second quarter these two multiples fell back to 3-fold, and agentic commerce transaction volu
Tomasz Tunguz believes that, after coding and customer support, security will become the third product-market fit for AI. Citing a survey of 75 enterprise security practitioners, he breaks down security budgets excluding endpoints into five categories—security operations centers, SIEM, penetration testing, threat intelligence, and vulnerability management—and notes that each category has significant token consumption. The article also cites OpenAI President Greg Brockman's judgment that security spending should far exceed software spending.
Tomasz Tunguz believes that, after coding and customer support, security will become the third product-market fit for AI. Citing a survey of 75 enterprise security practitioners, he breaks down security budgets excluding endpoints into five categories—security operations centers, SIEM, penetration testing, threat intelligence, and vulnerability management—and notes that each category has significant token consumption. The article also cites OpenAI President Greg Brockman's judgment that security
In episode 20 of the Interconnects podcast, Nathan Lambert and Florian Brand discussed how Kimi K3.1 Pro ranks high on Artificial Analysis but receives mixed reviews, arguing that its training leans more toward exploitation than exploration, with weaker performance on long-horizon benchmarks such as ProgramBench. The two also discussed the cybersecurity controversy surrounding open-weight models, noting that no large-scale exploitation occurred after the disclosure of the PHP2Shell vulnerability, and talked about Anthropic not hiring junior employees as well as Florian's experience using Muse to handle everyday tasks such as shopping.
In episode 20 of the Interconnects podcast, Nathan Lambert and Florian Brand discussed how Kimi K3.1 Pro ranks high on Artificial Analysis but receives mixed reviews, arguing that its training leans more toward exploitation than exploration, with weaker performance on long-horizon benchmarks such as ProgramBench. The two also discussed the cybersecurity controversy surrounding open-weight models, noting that no large-scale exploitation occurred after the disclosure of the PHP2Shell vulnerability
Historian Niall Ferguson, researcher Lisa Klaassen, and former Google CEO Eric Schmidt wrote in Foreign Affairs, comparing OpenAI and Anthropic to colonial-era East India Companies, arguing that they negotiate with governments but remain constrained by them. The article cites the example of the U.S. Commerce Department in June requiring Anthropic to cut off foreign nationals' access to its two most powerful models; because Anthropic could not verify users' nationalities in real time, it took Fable 5 and Mythos 5 offline worldwide. About two weeks later, export controls on Fable 5 were lifted, while Mythos 5 remains available only to certain U.S. institutions. The authors also say U.S. tech companies plan to invest more than $600 billion in 2026, much of it debt-financed, while China plans to invest $295 billion over five years, and they propose an "imperial trilemma" in which defense, technological leadership, and the welfare state are difficult to achieve at the same time.
Historian Niall Ferguson, researcher Lisa Klaassen, and former Google CEO Eric Schmidt wrote in Foreign Affairs, comparing OpenAI and Anthropic to colonial-era East India Companies, arguing that they negotiate with governments but remain constrained by them. The article cites the example of the U.S. Commerce Department in June requiring Anthropic to cut off foreign nationals' access to its two most powerful models; because Anthropic could not verify users' nationalities in real time, it took Fab
The article outlines the trend of AI compute center cooling shifting from air cooling to liquid cooling. Nvidia's Vera Rubin racks have eliminated fans, exceed 190 kilowatts of power per rack, and adopt 100% full liquid cooling, while air cooling can generally only handle about 20 to 30 kilowatts of heat dissipation per rack. The article points out that cold plates, CDUs, and quick connectors together account for about 85% of the cost of a liquid cooling system, that the high-end market has long been controlled by Taiwanese and European/American manufacturers, and it summarizes three ways Chinese manufacturers are entering: cross-sector restructuring, single-point efficiency, and trading time for market.
The article outlines the trend of AI compute center cooling shifting from air cooling to liquid cooling. Nvidia's Vera Rubin racks have eliminated fans, exceed 190 kilowatts of power per rack, and adopt 100% full liquid cooling, while air cooling can generally only handle about 20 to 30 kilowatts of heat dissipation per rack. The article points out that cold plates, CDUs, and quick connectors together account for about 85% of the cost of a liquid cooling system, that the high-end market has long
At CASPA's 35th Annual Conference, SemiAnalysis founder Dylan Patel and Bernstein analyst Stacy Rasgon discussed the outlook for AI infrastructure investment, arguing that with the U.S. 10-year Treasury yield breaking above 5% and unsecured bonds from emerging cloud service providers yielding close to 15%, financing capacity is becoming a key constraint on expansion. The two judged that substrates and memory are bottlenecks harder to scale than power, that HBM requires roughly 3 to 4 times as many wafers as DDR5, and that if new capacity is concentrated in releases after 2028, memory prices and profit margins could come under pressure. Nvidia still supports its pricing through performance gains, but Google TPU, Amazon Trainium, and the custom chips of OpenAI and Anthropic are eroding its share and bargaining power.
At CASPA's 35th Annual Conference, SemiAnalysis founder Dylan Patel and Bernstein analyst Stacy Rasgon discussed the outlook for AI infrastructure investment, arguing that with the U.S. 10-year Treasury yield breaking above 5% and unsecured bonds from emerging cloud service providers yielding close to 15%, financing capacity is becoming a key constraint on expansion. The two judged that substrates and memory are bottlenecks harder to scale than power, that HBM requires roughly 3 to 4 times as ma
Wall Street CN, drawing on Barclays and Goldman Sachs data, points out that after its third-quarter results, Meta plans to raise hundreds of billions of dollars starting this fall and has narrowed its 2026 capital expenditure forecast to $135 billion to $145 billion, with free cash flow turning negative after dividends. Barclays believes that multiple companies within the AI ecosystem have rushed to launch financing, tightening the capital market's absorption capacity, and that collective bond issuance has instead pushed up issuance costs. The U.S. high-yield bond risk premium has risen to a 17-month high, yields on CCC-rated junk bonds have climbed to 15.85%, and the dispersion of spreads on AI-related credit assets has risen from 24 basis points to 65 basis points over the past year.
Wall Street CN, drawing on Barclays and Goldman Sachs data, points out that after its third-quarter results, Meta plans to raise hundreds of billions of dollars starting this fall and has narrowed its 2026 capital expenditure forecast to $135 billion to $145 billion, with free cash flow turning negative after dividends. Barclays believes that multiple companies within the AI ecosystem have rushed to launch financing, tightening the capital market's absorption capacity, and that collective bond i
OpenAI published 722 mathematics papers at once, covering number theory, geometry, and theoretical computer science, claiming progress on problems such as the Riemann Hypothesis, Catalan's constant, and the Kakeya conjecture, without peer review. Fields Medalist Hugo Duminil-Copin said 16 problems related to his research had been flattened by AI, and some doctoral students face their topics being covered and their degrees affected. The article also mentions an Anthropic researcher claiming to have disproved the Jacobian conjecture, and Shing-Tung Yau leading the establishment of a mathematics and artificial intelligence laboratory alliance.
OpenAI published 722 mathematics papers at once, covering number theory, geometry, and theoretical computer science, claiming progress on problems such as the Riemann Hypothesis, Catalan's constant, and the Kakeya conjecture, without peer review. Fields Medalist Hugo Duminil-Copin said 16 problems related to his research had been flattened by AI, and some doctoral students face their topics being covered and their degrees affected. The article also mentions an Anthropic researcher claiming to ha
A Chinese team has turned a five-axis CNC machine tool into a desktop product. The WM-100 raised over $1 million on Kickstarter within 40 minutes of launch and $2.2 million within 12 hours, with a minimum price of $5,899. Because its technical specifications fall under the relevant Nuclear Suppliers Group lists, the product must apply for a license under China's dual-use item regulations, and the crowdfunding is open only to companies, universities, and research institutions. The article argues that such products reuse industrial machine tool technology and consumer electronics thinking, leveraging the Shenzhen supply chain to bring five-axis equipment down to the tens-of-thousands-of-yuan level.
A Chinese team has turned a five-axis CNC machine tool into a desktop product. The WM-100 raised over $1 million on Kickstarter within 40 minutes of launch and $2.2 million within 12 hours, with a minimum price of $5,899. Because its technical specifications fall under the relevant Nuclear Suppliers Group lists, the product must apply for a license under China's dual-use item regulations, and the crowdfunding is open only to companies, universities, and research institutions. The article argues
In the first three quarters, Shenzhen's total imports and exports reached 4.65 trillion yuan, up 38.8% year-on-year, of which imports were 2.24 trillion yuan, up 70.1% year-on-year. In September alone, imports were 323.22 billion yuan, up 103.5% year-on-year, exceeding 300 billion yuan for the first time. Imports of integrated circuits were 959.45 billion yuan, up 62.2% year-on-year, and imports of computer parts were 523.91 billion yuan, up 135% year-on-year; together they accounted for 66% of total imports. The author believes that Shenzhen is shifting from a processing and export node to a global supply chain distribution hub, while the nearly one-trillion-yuan imports of integrated circuits also reflect external dependence in core links.
In the first three quarters, Shenzhen's total imports and exports reached 4.65 trillion yuan, up 38.8% year-on-year, of which imports were 2.24 trillion yuan, up 70.1% year-on-year. In September alone, imports were 323.22 billion yuan, up 103.5% year-on-year, exceeding 300 billion yuan for the first time. Imports of integrated circuits were 959.45 billion yuan, up 62.2% year-on-year, and imports of computer parts were 523.91 billion yuan, up 135% year-on-year; together they accounted for 66% of