In partnership with
This issue contains affiliate links and two labeled sponsor slots. Editorial rankings are never sold.
Hey,
THE STORY OF THE WEEK
Stripe is buying OpenRouter for more than $7 billion. Five times its valuation from May.
OpenRouter does one thing: it decides which AI model handles your request, across 400+ models for around 8 million developers. That is it. That is the $7 billion product.
Choosing is now the business. Not building the model, not hosting it, just knowing which one to reach for.
Which is a strange week for us to ship what we shipped.
01 · We built a free one that shows its work
✦ NEW · THE STACK BUILDER
Pick your job. Answer five questions. Get a ranked stack.
64 jobs across 9 verticals: SEO, video, image, writing, audio, productivity, marketing, coding, design. Every recommendation is scored against 1,900+ hand-verified feature grades, and the reasoning is shown for each tool.
Free. No account. Rankings are deterministic and blind to affiliate status.
The honest version of why this exists: most tool directories answer "what is popular." That is a different question from "what fits my job," and the gap between them is where people waste money. So instead of ranking tools in the abstract, the Stack Builder asks what you are actually trying to do (get a new site ranking, turn long videos into short clips, keep source off other people's servers) and scores against that.
Where the OpenRouter comparison ends. A router optimises for cost and latency inside a black box. Ours optimises for fit and tells you why, in public, where you can disagree with it. Different product, same underlying truth: in 2026 the scarce thing is not capability, it is knowing what to pick.
It is live on the homepage and inside a growing set of guides. Break it and reply to tell me how.
The Ultimate Claude Code Guide to ship like Anthropic engineers
AI will write 90% of code by the end of 2026, and only 10% of developers will stay relevant.
The engineers in that 10% aren't smarter. They just know how to use AI.
We put together the exact playbook Anthropic engineers use to make sure you're in the top 10%.
Sign up for The Code and get access to:
The Ultimate Claude Code Guide 2026 — 50+ tips and tricks to code 5x faster
The Code newsletter — learn the latest AI tools, tips, and skills to code faster with AI in 5 minutes a day
The ToolChase site has no ads, no pop-ups, and no cookie wall. Two labeled sponsors here are what pay for that, which is also why our scores stay unbuyable. A click costs you nothing and keeps the whole thing running. Want a slot? Here's how it works.
02 · Deadline: August 26
⏳ SEVEN DAYS
o3 retires from ChatGPT on August 26, closing the 90-day sunset that started with the May 28 notice. The standalone DALL·E GPT is going too.
Two precision points, because both get reported wrong: these are ChatGPT-surface changes only, and the o3 notice states explicitly that the API is unaffected. The DALL·E retirement kills the standalone GPT, not image generation, which continues in ChatGPT Images.
If you have saved prompts, custom GPTs, or a team workflow pinned to either, that breaks next Wednesday. Ten minutes now beats a confused Thursday.
03 · Also this week
⚠ THE ONE THAT SHOULD WORRY YOU
An AI security tool wrote a security hole. Wiz Red Agent found that a GitHub Copilot Autofix patch to Snowflake's snowflake-connector-net repo, applied June 18, replaced a safe input pattern with raw string interpolation of a GitHub issue title. That opened a shell-injection hole. It was exploited within five days.
The tool whose entire job is closing vulnerabilities opened one, and the review process waved it through because the patch looked like a fix. If you have auto-merge enabled on any AI-generated security patch, this is your reminder that "AI reviewed it" is not the same as "someone reviewed it."
👦 ChatGPT shipped a teen version. A tailored experience for 13 to 17 year olds that blocks self-harm and romantic or sexual conversation, with age-prediction routing minors into it automatically. Parents get quiet hours and high-risk safety notifications. Whatever you think of age-prediction, this is the first big consumer AI product to treat under-18 as a separate design problem rather than a policy footnote.
📈 OpenAI is reportedly heading for an IPO as early as September, with an S-1 expected within weeks. That filing would expose ChatGPT's actual revenue, margins, and unit economics for the first time. Everyone quoting AI revenue multiples is about to get a reality check, in one direction or the other.
Full running log at AI Tools News 2026.
Learn AI in 5 minutes a day
You don't have to scroll every AI thread, track every new tool, or watch every demo.
The Rundown AI breaks it all down for you — the latest AI news, tools, and tutorials in one free 5-minute email every morning.
Trusted by 2M+ professionals at Apple, Google, and NASA.
04 · Three guides, matched to the news
💻 Best AI Coding Agents 2026
After the Copilot Autofix story, worth knowing which agents let you gate what they merge.
🎨 Best AI Image Generators 2026
With the DALL·E GPT retiring and Google's Imagen 4 endpoints closed, the image landscape just shifted under a lot of saved workflows.
🔍 Best AI SEO Tools 2026
SEO is the most-used vertical in the Stack Builder. This is the long-form version of what it recommends.
05 · Prompt of the week: the model router
If a routing layer is worth $7 billion, doing it by hand is worth ten minutes. Most people pick one model and use it for everything, overpaying on easy work and under-powering the hard stuff.
You're a model selection analyst. Here are the recurring tasks I use AI for, and the model or tool I currently use for each: [LIST: task → what you use now].
1. For each task, does it actually need a flagship model, or would a mid or budget tier do it just as well? Be specific about what I'd lose.
2. Which tasks genuinely justify the expensive tier, and why?
3. Where would a local or open-weight model work, given the data involved?
4. Give me a simple table: task, current choice, recommended choice, reason.
Flag clearly where you are uncertain rather than guessing, and do not recommend a switch that costs me more time than it saves.
THE TAKEAWAY
Capability stopped being scarce. Choosing well is the product now, and the only version worth trusting is one that shows its reasoning.
Go break the Stack Builder and tell me what it got wrong. Genuinely: the bad recommendations are the useful feedback, and I read every reply.
See you next Wednesday,
Best,
Emre
Building an AI tool? Submit it for a free editorial review. Read The State of AI Tools 2026: 724 tools analyzed, only 6% truly free.