Zendoric

Zendoric

AI analysis and perspective, every day.

Updated: September 22, 2026
🎙️ Listen to today's podcast (in Spanish)

🔥 In focus today

Sponsored Agents Make Ads Conversational — and Turn ChatGPT's Separation Wall Into the Real Product

🕒 Published on Zendoric: September 22, 2026 · 07:23

OpenAI is testing ads you can talk to: business-funded agents that pick up the conversation after a click, kept — the company says — separate from ChatGPT's own answers. The ad unit is no longer a banner, it's a persuasion system that argues back. Which makes one design detail load-bearing: whether that wall holds, and whether anyone outside OpenAI can check.

Read the analysis🎬 Video

📰 More AI analysis

Resy suspends a VC for using AI to book a table: the web still can't tell a legitimate agent from a scalper bot

🕒 Published on Zendoric: September 22, 2026 · 07:23

JC Bahr-de-Stefano connected the AI agent Instinct to his Resy account to land a highly coveted table in New York. The agent fired off 200 requests per hour and Resy suspended the account for behaving like a scalper bot: the web still can't tell a legitimate agent from a malicious one.

Read the analysis

UPB answers AI fears with a metaphor: the 'Intelligence of Love'

🕒 Published on Zendoric: September 22, 2026 · 07:23

UPB (Medellín) counters AI anxiety with the 'Intelligence of Love': a neologism, as its own rector admits, not a technical concept. Behind the metaphor lies an intuition that does hold up to scrutiny: judgment and human connection are a frontier the algorithm does not cross.

Read the analysis

Tracker false positive: 'AGI' is a gold miner, not artificial intelligence

🕒 Published on Zendoric: September 22, 2026 · 07:23

The article flagged under the 'superintelligence' topic has nothing to do with AI: 'AGI' is the stock ticker for Alamos Gold Inc, a Canadian mining company, and the story is a routine financial report on its share price.

Read the analysis

Why AI text shows: two rules for using it as an editor, not an author

🕒 Published on Zendoric: September 22, 2026 · 07:23

Developer Thomas Ptacek proposes a method for writing with AI without it showing: never accept its literal words and distrust its flattery, because readers detect artificial prose 'in parts per trillion'. His recipe: use the model as an editor, not an author.

Read the analysis

The weapon of mass deportation isn't advanced AI: it's merging every database

🕒 Published on Zendoric: September 22, 2026 · 07:23

DHS's use of AI in immigration cases grew 36% between December 2024 and January 2026, according to the American Immigration Council. Lawyers, the EFF and the ACLU document how cross-referencing federal, state and commercial data makes it possible to detain first and justify later.

Read the analysis

AI agents are already acting alone inside companies: cybersecurity demands they get their own identity and a kill switch

🕒 Published on Zendoric: September 22, 2026 · 07:23

AI agents chain actions together without human approval at each step, and traditional identity systems can't govern them. KuppingerCole, in a report sponsored by Ping Identity, proposes treating them as 'digital workers': their own identity, risk-based permissions and a kill switch.

Read the analysis

AI enters U.S. election polling: fear for jobs is growing, though no figures were published

🕒 Published on Zendoric: September 22, 2026 · 07:23

A polling column in The Washington Post reveals that two new surveys detect greater concern among Americans about losing their jobs to AI. The article itself doesn't provide the figures: here we explain why that silence is data too.

Read the analysis

Amid Helene's rubble, the White House picks AI as the symbol of educational rebuilding

🕒 Published on Zendoric: September 22, 2026 · 07:23

Melania Trump visited a school in Ashe County (North Carolina), hit by Hurricane Helene, to honor a teacher for an AI classroom project. The district received 1,500 AI literacy licenses and 100 robotics kits, courtesy of NVIDIA and StudyFetch.

Read the analysis

Brynjolfsson (Stanford): youth employment falls 19% where AI automates, not where it complements

🕒 Published on Zendoric: September 22, 2026 · 07:23

A Stanford study quantifies what until now was intuition: among 22- to 25-year-olds, employment falls 19% in the occupations most exposed to AI compared with less exposed peers. Economist Erik Brynjolfsson, who signs it, argues the correct unit of analysis isn't the whole job, but the task.

Read the analysis

Anthropic folds Cowork into Claude and launches Docs and Slides: the race for the AI office speeds up

🕒 Published on Zendoric: September 22, 2026 · 07:23

Anthropic is merging Claude Cowork into the standard chat and adding Claude Docs and Claude Slides in beta: documents and presentations without leaving the conversation. A move parallel to OpenAI's with ChatGPT and Codex that turns the assistant into a direct rival to Microsoft 365 and Google Workspace.

Read the analysis

U.S. bishops lobby Capitol Hill for the child safety bill on AI, and Congress leaves without voting on it

🕒 Published on Zendoric: September 22, 2026 · 07:23

More than a dozen Catholic bishops met with U.S. lawmakers on September 16 to call for AI guardrails and support for the Kids Online Safety Act. They did so the same day Congress adjourned without voting on it and with no plans to return until after the November elections.

Read the analysis

The UN puts its statistics at the service of AI agents, after a test revealed just 21% accuracy in the models

🕒 Published on Zendoric: September 22, 2026 · 07:23

The UN and Google are launching UN System Data Commons, a platform that lets AI agents query the UN system's official statistics in natural language. It arrives after a UNICEF benchmark —133,000 answers, six models— that found average accuracy of just 21.2%.

Read the analysis

The other AGI: Ag Growth International launches grain conveyors, not artificial general intelligence

🕒 Published on Zendoric: September 22, 2026 · 07:23

Canada's Ag Growth International (AGI) unveiled three grain handling machines at the 2026 Farm Progress Show in Iowa. Nothing to do with the AGI of artificial general intelligence: an acronym clash worth clearing up before drawing conclusions that don't apply.

Read the analysis

Ericsson forecasts 38% will delegate tasks to AI agents by 2030: the figure is an expectation, not a measurement

🕒 Published on Zendoric: September 22, 2026 · 07:23

Ericsson ConsumerLab calculates that 38% of smartphone users expect to delegate everyday tasks to AI agents by 2030, up from 4% today. The figure is an expectation stated in a survey, not a measurement of real adoption, and it's worth reading while knowing who signs it and why.

Read the analysis