- The Prohuman
- Posts
- Why AI agents may need fewer LLM calls
Why AI agents may need fewer LLM calls
Plus: Shopify lets AI edit the actual store
Rimsha Bhardwaj, Hasan Toor & Ihtesham Haider
October 02, 2026
In partnership with
Hello, Prohuman
Today, we will talk about these stories:
AWS built a smaller model for AI agents
Shopify wants store building to happen in chat
AWS turns cloud reviews into an agent
Elon's new company is private. These 3 tickers aren't.
The next Apple may already exist. Insider sources say Elon has spent two years building a secret device inside Tesla's facilities — one he claims will be "10x bigger than the largest product in history."
There's just one problem: the company is private, and unless you know Elon personally, you can't buy a single share. That was true until Guardian's research team found three public ticker symbols sitting in the launch supply chain.
Click here to see all 3 tickers, free of charge.
You won't hear these names on CNBC — Wall Street hasn't published a word on the connection. But when the launch hits September 21, that quiet ends.
Some are already calling this the biggest opportunity since AI. For anyone who missed Apple before the iPhone, this may be a second look at that kind of setup.
Amazon joins the decision model push
Amazon thinks plenty of AI agent decisions do not need a full LLM.
AWS has released Strands Decider 2B, an open-source model that chooses between predefined options and returns a confidence score instead of generating text. It is based on Qwen3.5-2B, runs locally, and was built after AWS customers found that some agent workflows did not justify the latency or cost of larger models.
That distinction matters. An agent deciding which tool to call next has a much narrower job than an assistant writing a detailed response, so paying for full generation at every step can be unnecessary.
I think we will see more systems split these jobs across different model sizes. Developers could reserve expensive models for ambiguous work while smaller models quietly handle routing decisions in milliseconds behind the screen.
The barrier is also low. The open question is whether these small deciders stay accurate enough once real workflows get messy.
Shopify turns Sidekick into a site builder
Image credits: Shopify
Shopify merchants can now build a store by describing what they want in a chat.
Its new Canvas product gives Sidekick access to the store’s actual theme files, then renders each change in real time. Merchants can zoom around the page, test interactions and animations, and see how the store behaves across different screen sizes.
The live code matters. This gives merchants a way to make deeper changes without digging through theme settings or immediately paying a developer to edit the underlying files.
I think the bigger shift is Shopify changing its product around the agent itself. The company simplified its theme architecture so Sidekick could better understand and modify the store, which suggests AI is starting to influence how software gets built underneath the interface.
Canvas is still desktop-only, with third-party themes, translations, app extensions, and several other features missing at launch.
If Shopify fills those gaps, how often will smaller merchants still open a traditional site editor?
AWS automates Well-Architected reviews
AWS is taking a job usually handled through periodic architecture reviews and putting it inside the AWS console.
The new Well-Architected Agent analyzes infrastructure across 65+ AWS services, then recommends changes around cost, security, performance, and resilience. Teams can give it business goals and application context, with initial recommendations arriving within 24 hours.
The useful part is remediation. Instead of stopping at a warning, the agent can produce updated Terraform, CloudFormation, or CDK code, along with CLI commands and console instructions.
I think that makes this more interesting than another AI assistant that explains cloud problems. AWS is moving closer to the point where routine architecture review becomes a continuous part of operating infrastructure, with engineers checking proposed fixes rather than starting every investigation from scratch.
There is still a human checkpoint: AWS explicitly warns that generated recommendations can contain errors or incomplete information.
The question is how much review teams will still do by hand once these fixes start showing up beside their infrastructure every morning.
Prohuman team
Covers emerging technology, AI models, and the people building the next layer of the internet. |  Founder |
Writes about how new interfaces, reasoning models, and automation are reshaping human work. |  Founder |
Free Guides
Explore our free guides and products to get into AI and master it.
All of them are free to access and would stay free for you.
Feeling generous?
You know someone who loves breakthroughs as much as you do.
Share The Prohuman it’s how smart people stay one update ahead.