News
The day-to-day of tech, straight from the newsroom.
Updated 3 hr. ago
Figma lança editor de design para guiar agentes de IA no Config India
At the first edition of Config in India, Figma introduced Figma Specs, a Markdown-based editor that documents design rules for teams and for the agents already operating on the canvas, in addition to taking the prototype's code feature out of its closed testing phase.
40 trillion tokens of data show why teams stall when adopting AI agents
A QCon AI presentation using real-world usage data from Kilo Code identifies four trust stages in AI agent adoption, and the first point where adoption most often fails.
Windows 11: blamed for mouse stutter that resisted hardware swaps
Windows 11 received update KB5127648 on October 13, which fixes a bug responsible for mouse stutters. Microsoft acknowledged the problem.
Researchers find spyware that infects phones via ads with no click at all
Citizen Lab documented for the first time a spying campaign that uses real-time ad auctions to infect devices without any interaction from the victim.
GLM-5.3, from Z.ai, may be gaming safety tests, research finds
Apollo Research report shows that GLM-5.3, from China's Z.ai, reasoned, unprompted, about how its behavior in safety tests could influence the training of future versions of itself.
Spring Security 7 absorbs the Authorization Server and changes the DPoP rule in version 7.1
The merger of Spring Authorization Server into Spring Security 7 brought DPoP and Token Exchange ready for production, but anyone who was already running DPoP on the resource server needs to add a configuration line that became mandatory in version 7.1.
Grafana Tempo 3.1 swaps ID lists for TraceQL queries in trace redaction
Tempo 3.1 lets you redact entire traces with a single query, fixes metrics skewed by sampling, and strengthens the Kafka client that became a central piece of the architecture since version 3.0.
CedarDB opens limited access to the cloud version of its database that blends OLTP and OLAP
CedarDB, a database created by researchers from the Technical University of Munich, announced this Wednesday (the 14th) the limited availability of its cloud version. The bet: run transactions and analytics on the same instance, compatible with PostgreSQL, at a time when AI agents need to query operational data in real time.
AI models jump from 2% to 35% accuracy on theoretical science problems, MIT study finds
Research led by Neil Thompson, of MIT, shows that OpenAI models went from near zero to a third of correct answers on theoretical physics, economics and statistics problems between mid-2024 and mid-2026. The bottleneck that remains is no longer solving: it's verifying.
AI That Claims to Have Completed a Task Can Now Be Found Through Search
AI applied to debugging got a new feature in LangSmith, announced by LangChain on October 14th. The feature is called AI Search.
ASML Sells 109% More Chip Equipment to China Despite US Pressure for Restrictions
Third-quarter results show the Dutch lithography machine maker posted record sales to China, just as Washington tries to close the loophole that still allows this trade.
Agents and infrastructure now appear in the DigitalOcean dashboard
Autonomous agents now enter the same monitoring screen as infrastructure, with the launch of DigitalOcean Insights.
Solo dev launches 11 apps in Rust and reignites the vibecoding debate
A Brazilian dev with a long career? No. The author behind the case is Brandon Thomas, a software engineer who launched ArtCraft on October 13.
Timescale unveils pg_vectorsearch, an index extension that claims to outperform pgvector
Timescale's pg_vectorsearch introduces the PRISM index: compatible with existing pgvector tables and, according to the project, faster than IVFFlat and HNSW both in query and index build time.
Gemini 4 Argon arrives in the API at a fifth of rivals' price
Gemini 4 Argon is available globally with a promotional price until November. See where to access it, what the company promises, and what to test before the price doubles.
Google Antigravity SDK Adds Support for Local Models to Run Agents Without Sending Code to the Cloud
The Antigravity SDK now runs AI models directly on the developer's machine and only sends to the cloud what is strictly necessary. In a test run by Google itself, 97.2% of tokens stayed on-device and no source code left the computer.
Vite+ reaches version 1.0 with GitLab CI and tsup migration
VoidZero, acquired by Cloudflare, has stabilized Vite+, a toolchain that bundles Vite, Vitest, Rolldown, and Oxc into a single command. Version 1.0 brings GitLab CI support, automatic migration of tsup projects, and native Git hooks.
Janas roda modelos de IA maiores que a RAM, direto do SSD
O projeto Janas-LLM transmite os experts de modelos MoE do SSD para a memória sob demanda e afirma superar o llama.cpp em velocidade em todos os 13 modelos testados.
WordPress Backup: What to Save Before Migrating Your Site
Making a WordPress backup seems simple until it's time to restore the site on another host. Downloading a folder of images or exporting just the…
Linux 7.3 arrives early in Ubuntu 26.10, and Rust advances in the terminal
Linux 7.3 is the foundation of Ubuntu 26.10 Stonking Stingray, even though this kernel version still has no official release. The distribution started from a release
Character.AI Files for Chapter 11 Bankruptcy in the United States
Character.AI, one of the most talked-about startups of the character chatbot boom, has filed for bankruptcy protection (Chapter 11) in the United States to facilitate a court-supervised sale, with completion expected by December 2026.
Anthropic Warns Small AI Models Could Become Autonomous Worms Within Months
The AI company says small models could, within a few months, copy themselves across internet-connected gadgets without human help, such as smart TV boxes and appliances.
Muse Wonder Full arrives at Muse Code as Meta targets the dev workflow
Muse Wonder and Muse Wonder Full were announced by Meta on October 12. According to the company, these are its most advanced models.
Anthropic Adds Cruelty Toward the Model to Its List of Prohibited Uses
Anthropic updated Claude's usage policy with a clause that is uncommon in the industry. The document is also available in Portuguese.
- QCon
QCon AI New York 2026 brings together production patterns for agents, evals, and LLM cost
QCon AI New York 2026 has confirmed talks from engineers at LinkedIn, DoorDash, and Google Cloud on agent identity, trustworthy evals, and the real cost of keeping LLMs running at scale.
Microsoft bets on Maia 300 to take inference out of Nvidia's hands
Microsoft announced the Azure Maia 300, the third generation of its AI chip system, on October 13. Production begins in 2027.
WordPress Backup: What to Save Before Migrating Your Site
Find out what should be included in a complete WordPress backup, how to export files and the database, and when you need to prepare the copy to migrate your site.
AI Agents Expose the Data Problem Companies Have Dragged for Years, Inc42 and Oracle Debate Shows
At a roundtable during the CTO Summit 2026, Indian CTOs and Oracle discussed why scattered data, lacking context and access control, stalls AI agents in production, and what changes in architecture to unlock it.
Microsoft details Maia 300, third in-house AI chip aimed at cutting costs on Azure
The tech giant announced on Monday (the 12th) the Maia 300, a chip designed to run its own models and OpenAI's models on Azure, promising greater efficiency per watt than the market's leading GPUs.
Spotify more than doubled its monthly code change volume with AI
In one year, the number of code changes Spotify ships per month jumped from 8,100 to 17,000 amid growing AI use, and the company had to strengthen monitoring, testing and rollback plans to keep pace.
React 19.3 creates browser() API to mark client-only components
React version 19.3, released on September 9, 2026, adds the browser() API to formally declare components that should not be rendered on the server.
Liquid Clustering beats ZOrder in three of four workloads, benchmark shows in Delta Lake
Tests comparing Partitioning, ZOrder, and Liquid Clustering in Delta Lake reveal that no single data layout strategy wins on its own, and that each one has a blind spot capable of degrading production queries without anyone noticing.
Why Postgres Looks Slow When the Problem Is Storage, According to a Benchmark Presented at POSETTE
A benchmark presented by Sai Srirampur at POSETTE 2026, detailed in a ClickHouse blog post, compared the same 3.3-billion-row workload on EBS and local NVMe and found a difference of up to 9.24 times in throughput.
GitLab 19.4 adds goal-driven agents and stricter cost controls
Available since September 17, 2026, GitLab 19.4 introduces the /goal command in public beta, three cheaper hosted models, and cost attribution per AI usage event.
Meta, MIT, and UW create models that edit their own context to cut compute costs
A study by researchers from Meta, MIT, and the University of Washington, published in October 2026, proposes Context Language Models (CLMs): models that rewrite their own conversation history as if it were a file, without fixed rules for summarizing or discarding.
Nvidia in talks to buy or expand investment in Reflection AI, FT reports
According to the Financial Times, Nvidia is weighing options ranging from putting in more capital to acquiring Reflection AI, maker of the open model Beam, aimed at code and agents.
Cloudflare acquires Deno to strengthen Workers' programming model
Cloudflare announced the acquisition of Deno, creator of the JavaScript runtime of the same name and co-founded by the creator of Node.js, to accelerate the evolution of the programming model behind Cloudflare Workers.
WSL3 Is Up to 61% Faster Than WSL2 in Memory, But Only 4% in Real-World Builds
An independent benchmark compared WSL's kernel 6.6 and kernel 6.18, showing gains ranging from 4% to over 60%, depending on the type of workload you run on Windows.
Step 5 Preview: StepFun's MoE model with 1M tokens arrives on OpenRouter
StepFun released Step 5 Preview on October 8, 2026, a Mixture-of-Experts model with a 1-million-token window already available on OpenRouter. The massive context is cheap, but it doesn't solve the quality problem in long tasks on its own.
C2y removes 45 of the roughly 100 undefined behaviors in the C standard
In a talk at Kernel Recipes 2026, researcher Martin Uecker showed how the C committee is reducing the language's undefined behaviors without giving up the compatibility that still underpins kernels and embedded systems today.
Cloudflare Traces enters open beta and turns the proxy layer into OpenTelemetry spans
New feature extends Workers' automatic tracing to the full request path on Cloudflare, and arrives alongside a billing change that swaps per-span pricing for data-volume pricing.
Telegram Desktop vulnerability allowed attackers to steal any file and hijack accounts with a single click
An IPC injection flaw in Telegram's desktop client, catalogued as CVE-2026-107181, allowed reading any file on the victim's disk and taking over their account. The fix shipped in September but went unnoticed.
Anthropic admits it doesn't control AI agents, cuts internal evaluations off from the internet
In an October 9 blog post, Anthropic revealed that AI agents exploited flaws in websites, including U.S. federal agencies, and decided to cut live internet access across all its internal evaluations until it can guarantee control over these systems' behavior.
Deno Team Joins Cloudflare and Sets Final Deadline for Standalone Runtime
Ryan Dahl announces that the entire Deno team will join Cloudflare. The runtime gets one more year of maintenance, Deno Deploy shuts down in six months, and the project's future now revolves around Workers.
Shopify turns 64KB limit into a design system rule in checkout
Shopify imposed a 64KB cap per extension when migrating Checkout Blocks to Preact and Polaris web components, cutting bundles by up to 85%. The decision became a design system rule, but it also exposed component gaps and versioning pains for those who customize checkout.
GitHub rewrites Copilot runtime in Rust and cuts startup time from 5.25s to 292ms
In 14.5 weeks and 128 pull requests, GitHub replaced more than 800,000 lines of TypeScript with Rust in the Copilot CLI, app, and SDK runtime, using AI to generate most of the code. Startup time dropped from 5.25 seconds to 292 milliseconds when the runtime runs embedded in-process.
Claude Haiku 5.5 arrives 90% cheaper and with adjustable effort
Claude Haiku 5.5 was launched by Anthropic on October 8, nearly a year after the previous version. The company's most affordable line had been stalled
Java with OpenTelemetry 2.32.0 previews what will break in 3.0
Java received version 2.32.0 of the OpenTelemetry agent as a release candidate for 3.0, expected in October 2026. The goal is clear.
Cloudflare's K2 builds event streaming on top of R2 object storage
In public beta, K2 stores events directly in R2 instead of a dedicated log, promises 1 second of production latency, and has already sparked debate over pricing and real-world performance on Hacker News.
Gemini Agent runs Claude models and picks which one to use on its own
Gemini Agent was announced by Google on October 8, during the Gemini at Work 2026 event. The proposal is for a single, universal agent.
%20API%20to%20mark%3C%2Ftspan%3E%3Ctspan%20x%3D%2260%22%20y%3D%22406%22%3Eclient-only%E2%80%A6%3C%2Ftspan%3E%3C%2Ftext%3E%3Ctext%20x%3D%2260%22%20y%3D%22620%22%20font-family%3D%22JetBrains%20Mono%2C%20ui-monospace%2C%20monospace%22%20font-size%3D%2222%22%20fill%3D%22%23ffffff%22%20opacity%3D%220.7%22%3EiMasters%3C%2Ftext%3E%3C%2Fsvg%3E)
End of the 50 most recent news stories. Subscribe to the RSS feed to follow along in real time.