😸 Claude Sonnet 4.5, Buy it in ChatGPT, Microsoft "Vibeworkers"

PLUS: You can now buy stuff directly in GPT?!

Welcome, humans.

Well, gaslighting with AI just hit a whole new level!

This husband updated their joint ChatGPT account to always take the husband’s side in all conversations (classic)…

Allegedly, OP says his wife has been using GPT as a marriage counsellor (we don’t recommend this use-case) so he thought he’d get even by updating the custom instructions (which she presumably doesn’t know exist?? Otherwise: BUSTED) to take his side in all “relationship advice” or marriage issue related chats.

If you are the wife (or anyone else reading this) and you don’t know: to set custom instructions, click on your name in bottom left corner > Personalization > Update the “Custom Instructions.” box. You can also toggle customization on / off, set a personality (“cynic”, “robot”, “listener” or “nerd”) and even select some of ChatGPT’s own qualifiers to add to your customization, like “chatty”, “witty” “straight shooting” or our personal favorite, “Gen Z” (lots more available if you click the “…”).

If your GPT-5 talks too much like a robot already, you might want to mess around with these settings and see if that helps! You can also set instructions on a project to project basis (our preference) to set different rules for different types of conversations. Have fun… just please don’t gaslight your partners with this newfound power kthxbye!

Here’s what happened in AI today:

Anthropic launched Claude Sonnet 4.5 and new tools for building AI agents.
You can now buy products directly inside ChatGPT with "Instant Checkout."
Microsoft added an "Agent Mode" to Copilot to automate full projects.
DeepSeek also dropped a new, more efficient AI model.

Advertise in The Neuron here

Claude Sonnet 4.5, Microsoft Vibeworker, “Buy it in ChatGPT”, and a surprise DeepSeek launch…

Y’know the famous saying, “there are years when nothing happens, and days when years happen?” In AI, there are days when nothing happens, and days when FOUR major companies all drop something that could each be its own main story, good night and good luck newsletter writers!

Well, this Monday was one of those days! Enjoy!

First up: You can now buy things directly in ChatGPT! OpenAI launched "Instant Checkout," letting U.S. users buy products from Etsy (and soon, Shopify) directly within a ChatGPT conversation. You can now go from asking for a product to purchasing it with a single tap, no browser needed. Built with Stripe, OpenAI claims results are ranked on relevance, not ads. Why it matters: With 700M weekly users, OpenAI is turning its chatbot into a massive e-commerce channel, creating a "super-app" that could directly challenge Google's and Amazon's dominance. Read the details here.

Next, Anthropic released Claude Sonnet 4.5, a new model that smashes records in coding benchmarks and can manage complex tasks for over 30 hours. Beyond the model, Anthropic shipped a full suite of new tools. For the first time, Checkpoints let you save and rewind progress in Claude Code, there’s a new native VS Code extension for Claude, and it can create files directly. Oh, also? Claude Sonnet 4.5 can now make Claude.ai’s website.

Related to Claude Sonnet 4.5? Microsoft launched its new “vibeworker.” Microsoft upgraded its 365 Copilot with “Agent Mode,” transforming it from a chatbot into an AI that handles entire projects. Instead of just asking for help, you can now delegate complex tasks. For example, tell it to “run a full analysis on this sales data” in Excel, and it will create new sheets, build charts, and check its work. And yes, Claude 4.5 is available in Copilot Studio. Read more here.

Our take on all this: Why did everything launch Monday? Was it all supposed to go live today, or did one thing trigger another thing? We’re not sure. We know some was planned, but others sprung up on us. What we will say is DeepSeek also launched v3.2, and whether it was intentional or not, the US labs succeeded at “flooding the zone” to bury this announcement (to the bottom of this article, anyway). Here’s a full news brief on THAT too.

FROM OUR PARTNERS

AI agents are here to stay. Is your company ready?

Learn why IDC research suggests early adopters of AI are gaining a competitive edge – and why your organization needs a clear strategy. Download the new IDC white paper and discover how to gain a market advantage that keeps competitors behind.

Download here.

Prompt Tip of the Day

Looking for a job? Just need to refresh your LinkedIn? Olivia Moore’s got you; she just shared a great series of tips from Google Gemini on how to use Nano Banana (their image editor) to create a brand new headshot. See below:

Here's what to include in your prompt:

Lighting: “Soft but dimensional studio lighting that creates a subtle catchlight in the eyes.”
Framing: Chest-up shot with space above the head, facing camera with a “confident and approachable” expression.
Styling: “Contemporary blazer with hair in a relaxed yet professional style, with crisp focus on individual strands.”
Background: “Neutral, solid-colored studio backdrop.”

Here’s the full prompt (G docs version): just copy and paste this into Gemini with a recent photo of you and try it out! Here is an example that is 100% Neuron approved.

P.S: If it doesn’t work the first time, try it a few times with different source photos.

Treats to Try

*Asterisk = from our partners (only the first one!). Advertise in The Neuron here.

*Murf AI turns your text into human-like speech in 200+ voices, so you can narrate presentations or create audiobooks without recording anything yourself.
n8n is like an open source Zapier for automating tasks—if you want to spend ~8 hours automating a task that would take you 20 mins (but like, you have to do that task every day, so its worth it lol), then use n8n (watch this).
DeepSeek-V3.2-Exp analyzes your long documents and conversations (like 300-page reports or entire codebases) by intelligently focusing on relevant sections instead of processing everything, cutting your API costs by 50%+.
Figma Make turns your design mockups into functional prototypes, and the Figma MCP server allows AI coding assistants to access your Figma designs' components, variables, and styling directly, so they can generate code that matches your design system, rather than just working from screenshots—here’s how to set up the MCP server.
Maximor’s agents integrate with your finance systems to automate reconciliations, close cycles, and produce audit-ready reports without manual spreadsheets (raised $9M).
GLM-4.6 is the new flagship model from Zhipu AI, featuring a longer context window and superior coding and reasoning capabilities that are competitive with leading models like Claude Sonnet 4.
Imagine with Claude creates fully interactive apps from your prompts, like a choose-your-own-adventure game or management dashboard, where Claude Sonnet 4.5 generates the interface and powers all the backend functionality in real-time (only available for the next 5 days only if you’re a Claude Max user).

Around the Horn

OpenAI is coming for Meta’s “Vibes” platform directly now! (article); Andrew also pointed out this will be out “in the coming days”

OpenAI wants 250GW of power by 2033, and The Information says it should end 2025 w/ ~2.4GW, which is up from 230MW from when it started 2024.
Meta entered early discussions with Google Cloud to potentially leverage Google's Gemini AI models for improving ad targeting across its platforms.
OpenAI is also on track to hit its projected revenue of $13B by the end of the year; in the first half of 2025, it earned $4.3B and is now generating ~$1B a month w/ $17B+ in the bank.
Anthropic planned to aggressively expand globally in 2025, 3xing its international workforce and quintupling its applied AI team as its revenue run-rate surpassed $5B.
OpenAI officially rolled out its new safety routing and parental controls to ChatGPT following suicide and violence incidents linked to problematic chatbot interactions.
CA Governor Gavin Newsom signed SB 53 into law, which launches CalCompute (a public cloud to spur AI innovation), requires large AI developers to publicly disclose safety plans and report incidents, and adds strong whistleblower protections for lab employees.
Researchers found that small improvements in a language model (like ChatGPT)’s single-step accuracy (so how accurate each step is in a series of steps) compound exponentially into dramatically longer task completion horizons (ex: GPT-5 executes 2.1K+ steps now vs Claude-4's 432), challenging the narrative of diminishing returns from scaling and suggesting current short-task benchmarks severely underestimate the economic value of continued AI investment (AI bubble who?? I don’t know her!).
Ethan Mollick argues AI can now do most tasks, but most jobs are made up of lots of tasks, so AI just shifts what you do in your job, not replace your job; therefore, the key area to focus on at work is “what’s worth doing” as opposed to being productive just for the sake of producing… or we’ll all drown in a sea of dreaded workslop.
If you’re deep into AI, this LoRA paper from John Schulman at Thinky (ex OpenAI CTO Mira Murati’s $10B startup) is for you. Also, since someone asked us, this paper tries to explain wtf “AGI” (artificial general intelligence) really is.

FROM OUR PARTNERS

Your time is too expensive to waste.

Most leaders lose 15 hours a week to the wrong tasks. The Delegation Guide and Worksheet helps you take them back. BELAY makes it possible with remote staffing solutions — U.S.-based, flexible, and personally matched to fit your business.

Download BELAY’s Free Guide!