Daily Tech News - 2026-02-19

The AI landscape is shifting from static text to cinematic motion as developers grapple with the boundaries of automation and underlying infrastructure.

Today’s headlines are dominated by a massive expansion in generative video capabilities. Google’s Gemini 3.1 Pro has arrived on AI Gateway with refined "medium-thinking" controls, but the real story is the multi-model video surge; Kling, Veo, Wan, and xAI’s Grok Imagine Video are all now accessible via unified APIs. This signals a move toward cinematic AI as a standard developer tool, a trend reinforced by the latest SWE-bench leaderboard where 4.5 Opus and Gemini 3 Flash continue to set new benchmarks for autonomous engineering.

On the infrastructure front, the push for efficiency is reaching the silicon level. NVIDIA is highlighting significant performance gains through NUMA node localization for its Blackwell and Hopper GPUs, ensuring hardware keeps pace with these escalating software demands. Meanwhile, Airbyte has launched its Agent Engine into public beta, providing the "connective tissue" for AI agents to interact with external platforms like Salesforce and GitHub. In a surprising pivot for systems programming, the Ladybird browser project has officially abandoned its adoption of Swift, highlighting the complexities of memory-safe language transitions in large-scale projects.

The human element of these tools is also coming into sharper focus. We’re seeing reports of "parallel agent psychosis," where developers lose track of code across ephemeral cloud instances—only to have AI agents extract and recover the "lost" work from session logs. As industry veterans like Eric S. Raymond note, the shift to AI-assisted programming isn't just about speed; it's about unlearning old habits. From Elastic’s new serverless capabilities to the refined design of Streamdown 2.3, the industry is racing to reduce friction in an increasingly automated ecosystem.

Featured Articles

accelerating-data-processing-with-nvidia-multi-instance-gpu-and-numa-node-localization-93421047

NVIDIA flagship data center GPUs in the NVIDIA Ampere, NVIDIA Hopper, and NVIDIA Blackwell families all feature non-uniform memory access (NUMA) behaviors. Most programs therefore do not have an issue with memory non-Uniformity. However, as bandwidth increases in newer generation GPUs, there are significant performance and power gains to be had.

Keywords: nvidia architecture, gpu memory, gpus numa, gpu instances, gpus nvidia, nvidia gpus, nvidia data, local memory, gpus consider, nvidia cuda
Source: developer.nvidia.com

access-billing-usage-and-cost-data-via-api-3298e04f

The /billing/charges endpoint returns data in the FOCUS v1.3 open-standard format. The endpoint supports 1-day granularity with a maximum date range of one year. Vantage has also released a native integration that connects Vercel teams to Vantage accounts.

Keywords: vercel billing, listbillingcharges, listbillingcharges 2025, billing listbillingcharges, billing, billing usage, billing charges, access billing, new billing, current billing
Source: vercel.com

airbyte-agent-engine-enters-public-beta-airbyte-59d80f9b

Introducing Airbyte Agent Engine. Every agent you build eventually needs to talk to the outside world. We're launching with 20+ agent connectors today, including HubSpot, Salesforce, Gong, Linear, GitHub, and more.

Keywords: agent infrastructure, agents need, external apis, enabling agents, repositories agents, agent needs, agent build, builds agent, make agent, vendor apis
Source: airbyte.com

catch-all-post-1468bc79

If you're trying to learn math with AI, don't try to guess how it should teach you. Here is Eric S. Raymond learning that he didn't miss what he thought he'd miss about the old way of programming. There are lots of posts of the form: "I just spent X hours cleaning up after AI"

Keywords: programming better, programming, learning technology, way programming, learning, learn math, enjoyment programming, programming asked, especially learning, python
Source: natemeyvis.com

gemini-3-1-pro-8c426e37

Gemini 3.1 Pro is the first in the Gemini series. It's priced the same as Gemini 3 Pro ( $2/million input,$ 12/million output under 200,000 tokens, $4/$ 18 for 200,00 to 1,000,000)

Keywords: svg pelican, pro gemini, gemini pro, improved svg, svg, preview gemini, svg animation, priced gemini, compared gemini, update gemini
Source: simonwillison.net

gemini-3-1-pro-is-live-on-ai-gateway-cb2a14fd

Gemini 3.1 Pro Preview from Google is now available on AI Gateway. This model supports the mediumthinking level for finer control over the trade-offs between cost, performance, and speed.

Keywords: ai gateway, preview ai, ai sdk, google gemini, streamtext ai, gemini pro, view ai, preview google, available ai, ai const
Source: vercel.com

go-modules-for-package-management-tooling-59ddc1ee

I’ve been working on a reusable layer for building ecosystem-agnostic package and supply chain tools in Go. I built the Go versions for git-pkgs, a tool for exploring the dependency history of your repositories that compiles to a single binary with no runtime dependencies. These are rebuilds of libraries I've written and used in Ruby for years.

Keywords: git pkgs, ruby libraries, dependencies git, dependencies maintainers, dependency history, dependencies purls, packages ecosystems, package rails, pkg repository, platforms rubygems
Source: nesbitt.io

grok-imagine-video-on-ai-gateway-eed7a1e8

Generate high-quality videos with natural motion and audio using xAI's Grok Imagine Video, now in AI Gateway. Try it out now via the v0 Grok Creative Studio, AI SDK V6 or by selecting the model in the AI Gateway playground.

Keywords: generatevideo ai, video generation, grok video, video generate, video ai, generate videos, generate video, generatevideo, generatevideo model, generatevideo import
Source: vercel.com

how-gnu-tar-handles-deleted-things-in-incremental-tar-archives-b169af6f

How to deal with deleted files is one of the challenges of incremental backups. Every incremental tar archive carries a full listing of the directory tree, covering both things that are in this incremental archive and things that come from previous ones. You can reconstruct the full directory tree as of that point in time from any single incremental archive.

Keywords: tar incremental, backup tar, incremental tar, tar archives, tar archive, gnu tar, incremental archive, file tar, incremental backups, incremental archives
Source: utcc.utoronto.ca

introducing-elastic-serverless-plus-add-on-with-aws-privatelink-capability-3fb85b31

Elastic Cloud Serverless enables organizations to build and scale faster, offering high agility and operational efficiency. With Serverless, teams can deploy their Elastic workloads faster and scale automatically without managing underlying Elasticsearch clusters or Kibana instances.

Keywords: elastic serverless, privatelink elastic, privatelink serverless, cloud serverless, elasticsearch serverless, serverless private, aws privatelink, aws private, elastic cloud, serverless capabilities
Source: elastic.co

is-the-future-aws-for-everything-8d0ac0a9

A theme running through my book is the idea that efficiency improvements have historically been dependent on some degree of repetition. It’s obviously hard to justify the time and effort it takes to design a really efficient production process if that process is constantly changing. An example of this playing out in practice is the different cost trends of cars vs. car repair.

Keywords: future aws, aws efficiently, efficient production, industries aws, aws benefits, higher production, aws concept, costs production, utilization aws, efficiency improvements
Source: construction-physics.com

kling-video-models-on-ai-gateway-2fdaff72

Kling models are known for their image to video models and multishot capabilities. Video generation is in beta and currently available for Pro and Enterprise plans and paid AI Gateway users. Kling 3.0 can generate multi-scene videos from a single narrative prompt.

Keywords: generatevideo ai, kling video, generate cinematic, video generation, generatevideo model, video generate, video models, klingai lastframeimage, cinematic videos, generate videos
Source: vercel.com

ladybirdbrowser-ladybird-abandon-swift-adoption-6adb3c44

In August 2024 the Ladybird browser project announced an intention to adopt Swift as their memory-safe language of choice. As of this commit it looks like they've changed their mind:Everywhere: Abandon Swift adoption.

Keywords: adopt swift, swift adoption, abandon swift, swift, 2026 ladybirdbrowser, 2024 ladybird, ladybird browser, ladybird abandon, swift memory, ladybird
Source: simonwillison.net

notes-about-this-blog-6debd28d

Everything on this blog that is not clearly marked as an LLM output (or other quotation) comes from me. I use AI for research, editing, thinking through questions, transcribing voice memos, and organizing thoughts into groups. I can't promise I'll never make a mechanical error and accidentally copy-paste an AI output in the wrong place.

Keywords: ai blog, eventually blog, blog, blog notes, blog se, generative ai, paste ai, blog rss, notes blog, include ai
Source: natemeyvis.com

recovering-lost-code-5a13fc1c

Reached the stage of parallel agent psychosis where I've lost a whole feature - I know I had it yesterday, but I can't seem to find the branch or worktree or cloud instance or checkout with it in. Turns out I'd been hacking on a random prototype in /tmp and then my computer crashed and rebooted and I lost the code... but it's all still there in ~/.claude/projects/session logs and Claude Code can extract it out and spin up the missing feature.

Keywords: openai mission, lost feature, agent psychosis, missing feature, parallel agent, agent, lost code, claude projects, evolution openai, code claude
Source: simonwillison.net

streamdown-2-3-refreshed-design-and-interactive-playground-523075fd

Streamdown 2.3 enhances design consistency by applying a unified nested-card design to tables, code blocks, and Mermaid diagrams. Action buttons now remain sticky during scroll, and code blocks render plain text immediately to reduce perceived latency before syntax highlighting loads.

Keywords: markdown new, streamdown enhances, markdown, streamdown, inconsistencies streamdown, syntax engine, code blocks, syntax highlighting, new interactive, markdown editable
Source: vercel.com

swe-bench-february-2026-leaderboard-update-9096fdf7

SWE-bench is one of the benchmarks that the labs love to list in their model releases. The official leaderboard is infrequently updated but they just did a full run of it against the current generation of models. 4.5 Opus is top, then Gemini 3 Flash, then MiniMax M2.5.

Keywords: benchmarks labs, bench benchmarks, benchmark uses, benchmark results, bash benchmark, benchmarks, benchmark, benchmark runs, api benchmark, good benchmark
Source: simonwillison.net

veo-video-models-on-ai-gateway-e0ecbdd6

Video generation is in beta and currently available for Pro and Enterprise plans and paid AI Gateway users. Try them out now via AI SDK V6 or by selecting the models in the AI Gateway playground. The Veo models are known for their cinematic quality and audio generation.

Keywords: generatevideo ai, generatevideo model, video generation, video generate, generatevideo, generate videos, generate cinematic, experimental_generatevideo generatevideo, experimental_generatevideo, await generatevideo
Source: vercel.com

video-generation-with-ai-gateway-c7aae85f

Video generation is in beta and currently available for Pro and Enterprise plans. Generate videos programmatically with the same interface you use for text and images. One API, one authentication flow, one observability dashboard across your entire AI pipeline.

Keywords: generatevideo ai, generatevideo model, generate videos, generate video, generatevideo, video generation, generates video, ai gateway, generation video, await generatevideo
Source: vercel.com

wan-models-on-ai-gateway-d5c80edf

Video generation is in beta and currently available for Pro and Enterprise plans and paid AI Gateway users. Try them out now via AI SDK V6 or by selecting the models in the AI Gateway playground. Wan produces artistic videos with smooth motion and can use existing content.

Keywords: generatevideo ai, generatevideo model, generate videos, video generation, video generate, generatevideo, video models, generatevideo import, v6 generatevideo, await generatevideo
Source: vercel.com