To everyone watching AI agents grow from chat windows into actual work systems: this week made the shift feel less theoretical.

OpenAI says agents are taking on longer delegated tasks, Grok 4.6 is built for long-running agent work, Vercel showed how its AI SDK team uses a software factory, GitHub brought Agent Plugins into Copilot surfaces, and Anthropic says Claude helped push forward a serious math result.

Also: the Toolfolio Discord is open. Share your tool, find early users, swap workflows, or hang out with other builders here: Join the Toolfolio Discord

Here’s what’s inside:

Top Story → OpenAI says agents are transforming work
AI Models → Grok 4.6 lands in Cursor and Grok Build
Dev Platforms → Vercel turns AI SDK maintenance into a software factory
Agent Tools → Agent Plugins 1.0 reaches VS Code, Copilot CLI, and the Copilot app
AI Research → Claude pushes deeper into math research

→ But first, check out today’s sponsor to support our newsletter  👇

Still using GoDaddy or Namecheap? There’s a better way with Porkbun!

Porkbun is the domain registrar trusted by creators, developers, entrepreneurs, and folks who want low prices without the nonsense.

Why people are choosing Porkbun:
• Most domains sold at cost
• Low, transparent registration and renewal pricing
• Free features like WHOIS privacy and SSL certificates
• Powerful web and email hosting options
• Real human support 24/7, 365 days a year
• Named the #1 domain registrar by Forbes Advisor and USA Today

For launching a business, building a personal brand, starting a side project, or creating your first website, Porkbun makes it easy.

Get $1 off your next domain registration with Porkbun now.

Get Your Domain Name Now

OpenAI published new research on how agents are transforming work, and the important part is not another flashy demo. It is the pattern in how people are actually using agents.

The research focuses heavily on Codex-style delegated work: tasks that run longer, involve more steps, and look less like “answer this question” and more like “take this workstream and move it forward.”

OpenAI says users are asking agents to handle more complex jobs over time, with non-developer usage growing especially fast. That matters because the agent market is no longer only about programmers telling a model to edit code. It is about legal, finance, recruiting, operations, and other teams handing off pieces of real work.

The bigger takeaway: the next AI interface might not be a better chatbot. It might be a work queue where agents accept goals, run for a while, and come back with finished output.

xAI released Grok 4.6, with a focus on long-running agents and more ambitious interactive and visual work.

The model is available now in Cursor and Grok Build, with xAI offering 2x included usage for the first week.

The positioning is very clear: Grok 4.6 is not only trying to be a chat model. xAI is talking about multi-step coding, topic research, visual tasks, web development, CAD-style work, self-testing, and turning broad ideas into workable projects.

For Toolfolio readers, the interesting part is where the model launched first. Cursor has become one of the main distribution surfaces for frontier coding models, which means developers may feel model shifts inside their editor before they feel them in a standalone chatbot.

Vercel published a detailed breakdown of the software factory it built for AI SDK.

The AI SDK has become a huge open-source project, and Vercel says its factory now uses specialized agents to classify issues, reproduce bugs, review PRs, handle backports, and move routine maintenance forward while humans stay in control of merges.

The headline number is wild: after four weeks, the software factory was authoring roughly 25-40% of merged PRs. Vercel also says the system resolves a large share of incoming issues and improves by learning from failed runs.

This is one of the clearest examples of agents becoming production infrastructure. Not a demo. Not a prompt thread. A real team using agents to keep a major developer tool moving.

GitHub announced Agent Plugins 1.0 support in VS Code, Copilot CLI, and the GitHub Copilot app.

Last week, Agent Plugins looked like an important packaging idea. This week, it became more practical: users can install plugins from the Awesome Copilot marketplace, and teams can start moving skills and MCP-backed tools across compatible clients instead of rebuilding the same integration for every surface.

The key idea is “build once, use across clients.” A plugin can bundle instructions, skills, and an MCP server into one portable package. Enterprise teams also get governance hooks through existing managed settings.

This matters because the agent ecosystem is getting messy fast. If every coding agent needs its own custom plugin format, teams drown in glue code. Agent Plugins is one attempt to make agent capabilities portable.

Anthropic published a research note on Claude’s mathematical capabilities, and it is one of the weirder, more interesting AI stories of the week.

An unreleased research version of Claude improved a longstanding lower bound related to the Riemann hypothesis: the fraction of zeros of the Riemann zeta function known to satisfy the hypothesis went from 41.6% to 67.2%.

Claude did not solve the Riemann hypothesis. That part matters. But Anthropic says the work happened across Claude Code sessions, used a huge amount of output, and produced a result that could be formally checked.

The practical takeaway is bigger than one math result. Agents are moving into research workflows where they can explore approaches, coordinate subproblems, write formalizations, and help humans test ideas that are too large for a normal chat exchange.

Learn more about Sponsoring Toolfolio here —> Toolfolio Boosts

Quicklinks:

Find the Latest Addition to Toolfolio here —> Newest Additions
Find our Dedicated Libraries of Tools here —> Libraries
20+ Coupons can be found on Toolfolio today —> View Coupon Codes

That’s all for this edition. Hopefully you found something new to spark ideas and make your workflow smoother.

Know a friend who’d love this? Forward it their way.

Got a tool worth featuring? Reply to this email

Catch you in the next drop,
— Dudu from Toolfolio