Good evening. Here is what matters in AI today, and how to put it to work.
Voters rejected AI infrastructure and PAC spending at the ballot box, while production agent adoption remains near zero: the gap between AI ambition and public reality is the story of the week.
~4 min read · last 12 hours
In today's issue
01
Voters dealt data centers a sweeping defeat in the 2026 midterms
02
Big Tech frames the AI backlash as partisan, not substantive
03
Uber fulfilled just 4 rides from AI agents in all of Q3 2026
04
Amazon Quick adds "Smart Approvals" to keep humans in the loop on agent actions
05
OpenAI acquires Pierre Computer Company to expand Codex's code tooling
Main story
Voters dealt data centers a sweeping defeat in the 2026 midterms
From Senate races to local rezoning votes, data center projects and their backers lost badly across the country, signaling a new era of community resistance to AI infrastructure.
Why it matters: Teams planning capacity expansion need to factor in local political risk as a first-class constraint, not an afterthought, before signing land leases or power agreements.
What to watch next: Watch whether state and local governments move to codify moratoriums into law following the ballot results, which would shift the data center siting challenge from political to legal and make it significantly harder to reverse.
The 2026 midterms delivered a clear signal: public tolerance for unchecked AI expansion, whether data centers in backyards or PAC money in campaigns, has run out, and the industry's attempts to spin the results away only deepen the credibility gap.
GPT-Image-2.6: Faster image generation with Flare and Sunburst
OpenAI
Inference Optimization with CoreWeave Mission Control Agent
CoreWeave
The Signal
The 2026 midterms handed the AI industry its clearest public rebuke yet, with data center projects and pro-AI political spending both losing decisively. The industry's instinct to explain this away as partisan noise is exactly the wrong read. For engineering and product leaders, the practical consequence is a tighter, slower, and more contested infrastructure environment, one where community opposition and regulatory friction are now real project risks. At the same time, the gap between agent hype and agent reality, illustrated by Uber's four rides, is a useful corrective: the engineering foundations for reliable agentic systems are still being built, and the teams investing in evaluation, context management, and human oversight now will be the ones with durable systems when adoption does arrive.
All the best, the KYFEX team
“Most of the risk is actually created by a couple of players building behind closed doors”
WIRED
Quick hits
Voters push back on AI infrastructure and political spending
Big Tech frames the AI backlash as partisan, not substantive
Tech industry spokespeople are attributing voter opposition to AI and data centers to anti-Trump sentiment rather than genuine concern, a framing most researchers and observers reject.
Why it matters: Misreading the cause of public opposition will produce the wrong strategic response and accelerate the case for hard regulation.
AI in production: agents, agentic gaps, and real deployment stakes
Across the stack this week, we see a consistent pattern: agentic AI is being built and funded aggressively, but actual production adoption is still thin, and the engineering work of evaluation, context management, and human oversight is what separates demos from durable systems.
Uber fulfilled just 4 rides from AI agents in all of Q3 2026
Despite enormous industry enthusiasm for agentic AI, Uber disclosed that only four rides were booked via third-party AI agents during its entire third quarter, a stark reminder that consumer agent adoption is still near zero.
Why it matters: Before investing in agent-facing APIs or integrations, product teams should validate that real users will actually invoke them, not just that the capability exists.
Amazon Quick adds "Smart Approvals" to keep humans in the loop on agent actions
Amazon Quick's new Smart Approvals feature uses a classifier to route only genuinely consequential agent actions to a human reviewer, reducing approval fatigue while preserving oversight where it counts.
Why it matters: This is a practical template for any team designing human-in-the-loop workflows: scope approvals by risk level, not blanket review, to keep agents useful without removing accountability.
OpenAI acquires Pierre Computer Company to expand Codex's code tooling
OpenAI is acquiring Pierre Computer Company, with the stated goal of expanding Codex's capabilities for building, understanding, and reviewing code.
Why it matters: The acquisition signals OpenAI's intent to own more of the developer workflow end-to-end, which matters for teams evaluating whether to build on Codex or maintain optionality with competing tools.
Learn to give AI agents the right information, in the right format, at the right time. In this course built with Chroma and taught by Jeff Huber, you'll develop practical skills for building...
Try out the features of the new Llama 3.2 models to build AI applications with multimodality.
DeepLearning.AI · Free · 1 hour
Put it to work
Try this today
Assess local regulatory risk for a new data center or AI infrastructure site
I am evaluating a potential site for AI infrastructure (data center, GPU cluster, or similar). The location is [CITY, STATE/REGION]. Please help me identify: 1) Any known local or state-level political opposition to data centers or AI infrastructure, 2) Recent ballot measures, zoning decisions, or moratoriums in this area, 3) Key community concerns (power draw, water use, noise, land use), and 4) Three specific due-diligence steps I should take before committing to this site. Be specific and flag any gaps in your knowledge that I should verify with local counsel.
Why it helps: Given this week's midterm results, infrastructure siting risk is no longer just an environmental review item, it is a business continuity question that belongs in every early-stage site evaluation.
Before you ship it
The risk
AI chatbots used for election or policy information can skew framing and act as an unaccountable intermediary between a citizen and factual sources, as researchers confirmed this week.
Do this
When deploying any AI tool that surfaces civic, regulatory, or policy information, ground every response in cited primary sources and route users to official government or legal references rather than letting the model summarize without attribution.
Ready to ship AI, not just read about it?
KYFEX designs and builds production AI for teams that need it working, not just demoed. Tell us what you're working on and we'll bring the engineering.
