Daily Tech News - 2026-02-17

The age of the autonomous agent has officially entered its high-growth phase, even as the "long tail" of software security casts a widening shadow. Leading today’s headlines is the release of Anthropic’s Claude Sonnet 4.6, featuring a massive 1M token context window and the ability to proactively delegate tasks to sub-agents. Alibaba joined the fray with the multimodal Qwen 3.5 series, further signaling a global shift toward agents that don't just process text, but navigate complex visual and tool-based environments with native ease.

This surge in capability is already transforming the developer workflow. Vercel Agent now provides automated build-fix suggestions directly within dashboards, while new tools like Showboat and Rodney are streamlining how agents document and test their own creations. For many, AI has become a "magic box" for rapid signal discovery, moving ideas from concept to prototype in hours by eliminating "infrastructure anxiety." However, this speed comes with a stark warning: Anthropic’s discovery of 500 critical vulnerabilities using Opus 4.6 highlights a massive backlog of abandoned open-source software that remains unpatched. In response, platforms like Skills.sh are pivoting toward automated security audits to maintain trust in increasingly agentic ecosystems.

On the consumer front, Apple’s iOS 26 adoption remains steady, matching the predictable rates of previous years as users focus on refining existing hardware controls. Meanwhile, the enterprise sector is looking toward NVIDIA’s new Multimodal RAG blueprint to ensure these powerful LLMs are grounded in trusted, visual-textual context.

Finally, amidst the digital acceleration, a nod to our roots and our world: we celebrate the anniversary of the first dial-up BBS going online in 1978, and conservationists are cheering the hatching of a rare kākāpō chick—a reminder that while our tools evolve at breakneck speed, some milestones remain timelessly significant.

Featured Articles

anthropic-s-500-vulns-are-the-tip-of-the-iceberg-b5214186

Claude Opus 4.6 can find critical vulnerabilities in established open source projects. Anthropic's 500 vulns are the tip of the iceberg. The scarier problem is the enormous long tail of abandoned software that nobody will ever fix.

Keywords: abandoned software, agent vulnerabilities, vulnerabilities pace, exploits agent, ai agents, vulnerabilities patches, vulnerabilities long, critical vulnerabilities, bugs maintained, critical bugs
Source: martinalderson.com

apple-releases-ios-26-adoption-rates-and-they-re-pretty-much-in-line-with-the-last-few-years-2c42d693

Apple has released its latest version of its mobile operating system, iOS 26. The update is expected to be pushed out to users in the next few weeks. The company has been slow to release updates to older versions of its software.

Keywords: ios 26, 26 ipados, 53 ipads, versions ios, ipados 26, released ios, ipads introduced, ios 18, ipados 18, rate ipads
Source: daringfireball.net

automated-security-audits-now-available-for-skills-sh-a645ed5e

Skills.sh now have automated security audits to help developers use skills with confidence. Security audits appear publicly on each skill's detail page. Skills flagged as malicious are automatically hidden from the leaderboard and search results.

Keywords: skills sh, skills counting, use skills, skills skills, protection skills, adding skills, skills, skill page, 000 skills, counting skills
Source: vercel.com

automatic-build-fix-suggestions-with-vercel-agent-bd701b92

Vercel Agent now scans deployments for build errors. When it detects failures it automatically suggests a code fix based on your code and build logs. It can also suggest code fixes inside the Vercel dashboard whenever a build error is detected.

Keywords: builds vercel, vercel agent, reviews vercel, vercel dashboard, agent build, suggestion github, started vercel, vercel, broken builds, agent reviews
Source: vercel.com

build-ai-ready-knowledge-systems-using-5-essential-multimodal-rag-capabilities-5028ad37

Retrieval-augmented generation (RAG) was created to ground LLMs in trusted enterprise knowledge. RAG must understand both visual and textual context to achieve enterprise-grade accuracy. The NVIDIA Enterprise RAG Blueprint is built for this. It connects unstructured enterprise data to the intelligent systems built on top of it.

Keywords: ai documents, retrieval metadata, reasoning metadata, ai blueprints, ai storage, embedded metadata, data enterprise, enterprise data, intelligent document, data ai
Source: developer.nvidia.com

claude-sonnet-4-6-is-live-on-ai-gateway-5dc3428f

Claude Sonnet 4.6 from Anthropic is now available on AI Gateway with the 1M token context window. The model proactively executes tasks, delegates to subagents, and parallelizes tool calls, with MCP support for scaled tool use.

Keywords: ai gateway, sonnet ai, ai sdk, available ai, streamtext ai, view ai, ai, learn ai, improvements agentic, adaptive ai
Source: vercel.com

first-bbs-goes-online-feb-16-1978-9889e4da

Ward Christensen and Randy Suess launched the first public dialup BBS in 1978. It operated over an analog phone line and a Hayes dialup modem running at 300 bits per second. They called the first BBS program CBBS, which stood for Computerized Bulletin Board System.

Keywords: dialup bbs, bbs program, program cbbs, called bbs, bbs bulletin, cbbs based, copies cbbs, bbs, byte magazine, ran cbbs
Source: dfarq.homeip.net

first-k-k-p-chick-in-four-years-hatches-on-valentine-s-day-79a4a57b

First kākāpō chick in four years hatches on Valentine's Day. First chick of the 2026 breeding season! Kākakapō Yasmine hatched an egg fostered from Tīwhiri. Second chick hatched this morning. This is Ako's first-ever chick.

Keywords: kakapo2026 breeding, kākāpō chick, kakapo chick, fostered kākāpō, kākāpō mums, second kakapo, kākāpō tīwhiri, chick hatched, chick kakapo2026, ako nest
Source: simonwillison.net

how-should-less-technical-people-get-started-with-agentic-ai-tools-8f2995b3

Don't follow a tutorial, for these reasons: Tutorials put you in a passive frame of mind. Start small and specific: instead of "automate my invoice creation," try "extract information X, Y, and Z from this folder of .pdf files"

Keywords: suggest starting, ai making, ai choice, tutorials, agentic ai, automate, started claude, especially automating, tutorial quickstart, started agentic
Source: natemeyvis.com

how-stably-ships-ai-testing-agents-in-hours-not-weeks-a58df445

A 6-person startup learned their bottleneck wasn't AI. It was infrastructure anxiety. Then they won Vercel’s AI Accelerator. Their shift highlights how Vercel's platform eliminates infrastructure anxiety, boosting autonomous testing and enabling quick enterprise growth.

Keywords: vercel ai, agents faster, ai infrastructure, builds ai, testing agents, stably ai, ai agents, devops overhead, ai scalability, agents vercel
Source: vercel.com

how-to-force-restart-an-iphone-e5d7d95a

Modifying this control will update this page automatically. Use the iPhone User Guide to help you get the most out of your iPhone. This guide includes instructions for iPhone models compatible with iOS 26. It also includes information on how to use the Camera Control and the Lock Screen.

Keywords: ios 26, 26 iphone, iphone 13, generation iphone, new ios, iphone 14, modification ios, iphone change, iphone 15, control iphone
Source: support.apple.com

i-swear-the-ufo-is-coming-any-minute-b8eaf514

When Prophecy Fails is supposed to be a classic case study of cognitive dissonance. The neurologist Oliver Sacks appears to have greatly embellished or even invented his case studies. A recent failure to replicate a basic dissonance effect, things aren’t looking great for the phenomenon.

Keywords: dissonance ufo, ufo cult, ufo coming, cult predicts, science prophecy, swear ufo, prophecy fails, fails prophecy, ufo, cult meetings
Source: experimental-history.com

improved-streaming-runtime-logs-exports-6e1aa1b3

Exports now stream directly to the browser. You can export exactly what's on your screen or all requests matching your current search. Exported log data is now indexed by request to ensure consistency with the dashboard interface.

Keywords: export logs, exports stream, dashboard export, logs exports, requests export, view export, export limits, exported log, export runs, export
Source: vercel.com

microspeak-escrow-25a26d98

As a product is nearing release, the release management selects a build and declares it to be the escrow build. The metaphor is that this build has been placed into the hands of an imaginary third party.

Keywords: term escrow, escrow phase, escrow terms, escrow period, described escrow, definition escrow, escrow build, period escrow, escrow, new escrow
Source: devblogs.microsoft.com

platform-strings-68b0e3c9

The format cpu-vendor-os dates to the early 1990s. LLVM inherited the format through Clang’s cross-compilation support. The architecture ARM calls “AArch64” is what Apple calls ‘arm64’

Keywords: mac compiling, compiler targets, compilation targets, binaries compiled, platforms gcc, compiled cpu, compiler apple, compiler target, x86 compiler, compiler metadata
Source: nesbitt.io

quoting-dimitris-papailiopoulos-6c0b5509

I now have something close to a magic box where I throw in a question and a first answer comes back basically for free, in terms of human effort. Before this, the way I'd explore a new idea is to either clumsily put something together myself or ask a student to run something short for signal. That quick signal step, i.e., finding out if a question has any meat to it, is what I can now do without taking up anyone else's time. It's now between just me, Claude Code, and a few days of GPU time.

Keywords: running research, research long, taking time, research questions, human effort, research, gpu time, explore new, questions, explore
Source: simonwillison.net

quoting-rough-draft-8-2-66-c860f29d

This is the story of the United Space Ship Enterprise. Assigned a five year patrol of our galaxy, the giant starship visits Earth colonies, regulates commerce, and explores strange new worlds and civilizations. These are its voyages... and its adventures.

Keywords: starship, ship enterprise, starship visits, giant starship, star trek, space ship, trek opening, voyages adventures, new showboat, patrol galaxy
Source: simonwillison.net

qwen3-5-towards-native-multimodal-agents-8798eebb

Alibaba's Qwen just released the first two models in the Qwen 3.5 series - one open weights, one proprietary. Both are multi-modal for vision input and output. The open weight one is a Mixture of Experts model.

Keywords: qwen3 native, models qwen, qwen3, qwen3 plus, use qwen, called qwen3, confusing qwen, qwen, interesting qwen, qwen just
Source: simonwillison.net

recraft-v4-on-ai-gateway-0bd872c6

Recraft V4 is now available on AI Gateway. It's a text-to-image model built for professional design and marketing use cases. The model has improvements with photorealism, with realistic skin, natural textures, and fewer synthetic artifacts.

Keywords: generateimage ai, generateimage model, produces images, await generateimage, import generateimage, generateimage, image model, ai sdk, ai gateway, v4 ai
Source: vercel.com

rodney-v0-4-0-fa0c0375

Rodney v0.4.0.0 has been released. It's a command-line tool for browser automation. There are a number of new features, including a command to test web apps in multiple steps.

Keywords: browser automation, running chrome, tool browser, browser window, make browser, chrome instance, tests run, browser, test web, cli tool
Source: simonwillison.net

rumors-of-agi-s-arrival-have-been-greatly-exaggerated-69a9be04

Rumors that humanity has already achieved artificial general intelligence (AGI) have been greatly exaggerated, say Gary Marcus and colleagues. They argue that recent claims of achieving AGI rest on a conceptual error: conflating statistical approximations with intelligence itself.

Keywords: approximations intelligence, intelligence agi, general intelligence, intelligence benchmark, intelligence general, agi benchmark, intelligence robustness, estimates ai, agi greatly, intelligence recent
Source: garymarcus.substack.com

the-new-try-yarn-spinner-3c5ea37c

We recently released a brand new Try Yarn Spinner, which I wrote. I also wrote a new demo story too! Check it out below:Technical post on the new Try yarn spinner coming soon.

Keywords: yarn spinner, spinner wrote, spinner coming, details yarn, try yarn, spinner, yarn, demo story, spinner january, new try
Source: hey.paris

two-new-showboat-tools-chartroom-and-datasette-showboat-7dca1618

Showboat is a tool that helps coding agents create Markdown documents that demonstrate the code that they have created. Chartroom is a CLI charting tool that works well with Showboat. Datasette-showboat lets Showboat’s new remote publishing feature incrementally push documents to aDatasette instance.

Keywords: showboat chartroom, showboat database, showboat datasette, database showboat, datasette showboat, showboat commands, showboat document, showboat plugins, showboat tools, showboat documents
Source: simonwillison.net

understanding-the-limitation-of-do-in-new-frame-window-in-gnu-emacs-7a57b815

GNU Emacs has a core model for how it operates, and some of its weird seeming limitations are easier to understand if you internalize that model. In order to create a frame, you must have some buffer to display on that frame. The code basically assumes you're running a command that will at some point try to display a buffer, and it hooks into that operation.

Keywords: emacs frames, command frame, frame gnu, frame operation, folder frame, frame folder, frame window, frame run, context frame, frame prefix
Source: utcc.utoronto.ca

vercel-sandbox-snapshots-now-allow-custom-retention-periods-3bc22c08

Snapshots created with Vercel Sandbox now have configurable expiration, instead of the previous 7 days limit, along with higher defaults. The expiration can be configured between 1 day to infinity. If not provided, the default snapshot expiration is 30 days.

Keywords: snapshot expiration, sandbox snapshot, vercel sandbox, configurable expiration, sandbox vercel, default snapshot, snapshots created, expiration configured, expiration, expiration 30
Source: vercel.com

weekly-update-491-743b28c1

The ESP32 Bluetooth bridge experiment was a complete failure. At a guess, BLE is a bit too passive to detect state changes. Unless it was awake and communicating, it just had no idea what was happening with the locks.

Keywords: yale locks, esp32 bluetooth, locks reliably, bluetooth bridge, locks actually, bluetooth, locks, happening locks, hope locks, lock related
Source: troyhunt.com