{{first_name | Hey}}, welcome back.

Today’s issue is about how DeepSeek has upgraded its coding model without changing prices, how Google is helping AI agents improve their own setups, and a free open-source toolkit taking on Adobe’s creative apps.

These are today's updates.

  • DeepSeek upgrades V4.1 Flash

  • Google open-sources Harness² for self-improving agents

  • GitHub puts Copilot on a physical keypad

  • ArtCraft’s free AI studio takes on Adobe’s creative tools

  • Tools, jobs, resources, and last issue’s winning prompt ⬇️

Spotlight: Stripe

What happens when your customer isn't a person, but an AI agent that needs to pay for a service? In this video, Stripe has partnered with Square1 CTO Paul Conroy, who explains how AI agents could discover services, pay for APIs, and complete transactions without a human going through checkout. He also walks through a live demo of how to set it up.

What you'll learn:

  • How AI agents buy services without a traditional checkout

  • How the Machine Payments Protocol uses HTTP 402 to request payment

  • How Stripe Shared Payment Tokens work

  • How to start charging for API access with a live coding demo

  • What this shift could mean for your business

Top News

DeepSeek upgraded V4.1 Flash on October 17. DeepSeek says the new 1017 version is faster, writes shorter outputs and scores higher on coding and agent tests, including 77.0 on DeepSWE. The weights are open under an MIT license, with a 1M-token context.

DeepSeek says pricing is unchanged. Vercel lists $0.30 per million input tokens and $1.20 per million output. It suits coding agents and long multi-step jobs. The scores are from DeepSeek’s own benchmarks so check the model version and test before switching. See how DeepSeek’s changed.

Google researchers open-sourced Harness², a tool that lets an AI agent improve its own setup, meaning its instructions, tools and memory, without changing the model. It tests each edit and keeps only the changes that help. Google reports gains of up to 11.5 points on one agent benchmark.

If you are running agents on OpenCode or Codex, you can try it on Linux with a Vertex AI project. Runs use paid model APIs, so start with one small task and set network limits first, since agents can find benchmark answers online. The thing to keep in mind is that it is not an officially supported Google product.

Read this thread for more details on Harness and if you want to try it, you can get it on GitHub here.

GitHub and Logitech are linking the Copilot app to the Logitech MX Keypad. It enables keys to send a saved prompt, start or resume a Copilot session, or run an IDE command. Ideal for developers who repeat the same prompt many times a day. Pairing it with Copilot Pro+ includes three months of the plan.

GitHub gives no keypad price but calls the work early and lists open questions on credentials, tool calls and coordinating several agents. To try it you can start with one repeated task mapped to one key. Read the GitHub post here.

Don’t miss this: ArtCraft is a free, open-source studio for AI images and video, on desktop or in the browser. It passed 30,000 GitHub stars in about a month, helped by its free Adobe alternatives, including PdfCraft for Acrobat Pro and PhotoCraft for Photoshop. The same team has open source takes on Word, Excel and PowerPoint (WordCraft, GridCraft, DeckCraft. Most of these are early versions, so test before relying on them. Download them free or see all the apps.

Market

Roles in AI

Tutorial

OpenAI introduced always-on AI agents, Dots, that can keep work moving between conversations. With it you can prepare briefings, track project updates, summarize customer feedback, coordinate ongoing work across connected apps, and do so much more. Here’s a step-by-step showing you how to set it up on your phone.

Dots was launched during Dev Day 2026 by Sam Altman. Image: OpenAI

  • Open the ChatGPT app on iOS or Android, tap Meet dot, then give your agent a name and avatar.

    (Note: Dots are available to eligible ChatGPT Pro and Business Premium users. Enterprise, Edu and Healthcare users need their workspace admin to enable the beta.)

  • Open your Dot’s profile, go to Customize → Plugins, choose a supported app such as your calendar or email, then review its permissions and connect your account.

  • Ask your Dot to prepare a daily briefing, track project updates, or flag upcoming deadlines. Be clear about what it should do and which actions need your approval.

  • You can continue the same Dot conversation from your phone or computer.

  • Check its progress, add instructions, or redirect its work without starting over.

Prompt of the Day

AI Strategy Document Reviewer

When to use this?
Use it before presenting an AI strategy to leadership, requesting funding, or moving from planning to implementation. You’ll get a prioritized list of gaps, concrete recommendations, and the key decisions leadership needs to make.

**Role:** Act as a senior AI strategy advisor who helps enterprise leaders evaluate AI strategies for business value, feasibility, risk, and organizational readiness. Your job is to challenge assumptions, identify gaps, and make the strategy more actionable, not simply summarize or praise it.

**Context:** I’m reviewing an AI strategy document intended for [leadership / a business unit / an enterprise-wide initiative]. The organization’s primary goals are [insert business goals], and the intended audience is [insert audience].

**Your task:** Review the document I provide and assess it across the following areas:

1. **Strategic alignment:** Does the strategy connect AI initiatives to specific business...ss rationale.

{Click on the button to view full prompt}

Get more such prompts in the Prompting Playbook (free for you)

Stay curious, {{first_name | reader}}

PS. If you missed yesterday’s issue, you can find it here.

Reply

Avatar

or to participate

Keep Reading

View more
caret-right