THE PERSONAL DAILYMonday, October 12, 2026Issue 007 / Edited edition
From your following Edited for reading
X PaperREAD
Your following. Your daily paper.
Take time to read See who said what
81 followed sources · This issue 9 storiesNamed voices · Traceable sourcesLatest issue
IN THIS ISSUEOCT 12
Politics & World
After the diesel license, the dispute shifts to refineries and an energy ceasefire
BBC News 中文 · @bbcchinese
BBC Chinese reported on October 11 that the US–Russia diesel deal drew Ukrainian and European opposition while Trump urged Ukraine to stop attacking Russian refineries. Issue 006 covered the license; this follow-up concerns political reactions, not an agreed energy ceasefire.
According to the BBC, Zelensky warned that easing diesel sanctions would fund Russia’s war. Trump linked refinery attacks to global fuel supply and called for new Ukrainian leadership. That is his political demand, not a Ukrainian decision to hold elections or replace the president. Elections remain suspended under martial law.
Britain called for an immediate halt to attacks on energy facilities and Black Sea shipping. EU foreign-policy chief Kaja Kallas opposed easing pressure on Russia and said Europe would not follow. Refineries, energy facilities and shipping are different scopes; these statements do not establish a mutually agreed ceasefire.
The report describes planned deliveries of 300,000 tonnes first and 500,000 in November, followed by one million and three million tonnes. Trump also made later delivery conditional on refinery conditions. Planned volumes are not completed shipments. We read the full BBC report but have no evidence of each delivery; actual price effects require subsequent data.
Tibo’s quota poll separates more capacity from uninterrupted work
Tibo · @thsottiaux
On October 11, Tibo asked Pro users to choose between five-hour limits with twice the weekly usage, and no five-hour limits with unchanged usage. The final poll showed 240,643 votes: 33.2% for the first option, 37.8% for the second and 29% for viewing results only.
“2X week usage” means twice the weekly allowance, not two uses per week. The distinction is between total capacity and the freedom to work continuously. Removing five-hour limits received more votes than the alternative, but not a majority of all votes.
This was a public preference poll, not an announced OpenAI plan change. Participation was self-selected, so the result cannot represent all paying users or promise a particular future quota.
A reported EU hybrid-car plan specifies 350,000 vehicles
StockMKTNewz · @stockmktnewz
An October 11 Bloomberg report added detail to earlier references to roughly 400,000 Chinese hybrid cars. Citing a document it had seen, it reported a first-year limit of 350,000, rising 6% annually from year two, for four years, potentially starting December 1.
The report, republished by Moneycontrol, says the cap would cover hybrid imports from all origins for non-discrimination, with China’s allocation no smaller than any other origin’s. This concerns hybrids, not an interchangeable rule for battery-electric tariffs.
We have not obtained an enacted legal text. The newer reported terms clarify the proposal; they do not establish that the cap is already in force. Final quantities, timing and scope still require the official instrument.
USITC opens a power-delivery patent inquiry; no infringement ruling yet
外汇交易员 · @fxtrader
The US International Trade Commission announced investigation 337-TA-1526 on October 9, covering vertical power-delivery systems, components and computing systems containing them. Vicor’s September 9 complaint alleges patent infringement involving imports and sales. Foxconn Industrial Internet and FII USA are among the respondents.
Vicor requested a limited exclusion order and cease-and-desist orders; those requests are not granted remedies. USITC explicitly said institution decides nothing on the merits. An administrative law judge will conduct an evidentiary hearing and issue an initial determination subject to Commission review.
The stated 45-day period concerns setting a target completion date, not finishing the investigation. This report establishes the procedural step, not infringement, a sales ban or the business impact claimed elsewhere.
Saudi Arabia’s civil aviation authority said King Khalid International Airport in Riyadh was attacked again on Sunday afternoon, October 11. The statement, carried by the Saudi Press Agency, reported injuries and ongoing damage assessments and facility safety inspections.
This Sunday statement supplied no injury count and named no attacker. Earlier reports circulated by a followed source concerned Saturday; their casualty figures must not be assigned to this separate attack.
The confirmed points are a further attack and continuing inspections. Flight restoration, the extent of facility damage and casualty totals for this incident require subsequent official information.
Karpathy sees a widening gap in how people experience AI
Andrej Karpathy · @karpathy
On October 11, Andrej Karpathy revisited an April observation: people discussing AI may be describing very different tools and tasks. He framed the gap as a funnel; Baoyu subsequently shared his account.
His layers run from people with little exposure and casual free-chat users to mathematics and coding professionals using frontier models, then a small group with internal access. He explicitly describes the population figures as rough estimates. They are not survey results, and the internal capabilities he describes have not been independently verified for this article.
He identifies three drivers: the ambition, size and duration of a task; stronger but uneven capabilities in digital, verifiable domains; and differences between free, paid and internal access. In his view, modest benefits for many users can coexist with rapid changes for some professionals.
This is an interpretation of divergent experiences, not evidence that every occupation will improve equally or that one group is behind. Readers still need to evaluate their own tasks, available systems and ability to check outputs. The April post supplies background rather than a new event.
Tibo announces 500k default context, with no long-context fee against subscription usage
Tibo · @thsottiaux
At 07:45 Beijing time on October 12, Tibo announced that GPT-6.1 Sol’s default context length in Codex and Work had increased from 272k to 500k. He then added that long-context fees would not be charged when usage is drawn from a subscription.
The change concerns the default setting. The model’s maximum context remains 1M, and users can override the default in either direction. Moving from 272k to 500k is an increase of about 83.8%, described by Tibo as roughly 80%. Context capacity concerns the information available to the model during processing; a larger default does not itself double subscription allowances or output length.
Tibo said the setting had been tested and was intended to reduce compactions and help tasks that need to retain substantial context and nuance over time. These are his product claims. This issue has not compared performance before and after the change, and the announcement does not say that compaction has been eliminated.
The fee statement specifically refers to usage drawn from a subscription. It does not announce free API usage, unlimited allowances or removal of the five-hour limit. His previous day’s poll about Pro allowances remains a separate consultation, not an implemented policy bundled into this update.
REA in practice: an annotation entry point, not automatic messaging
Baoyu · @dotey
At 06:53 Beijing time on October 12, Baoyu described using REA to examine annotations in the Codex/ChatGPT app. He wanted his own page to append a comment to the composer, helped by sending one message, then asked the agent to build a test page.
He reports that the test could open a comment editor and prefill content, but could not append the comment directly to the composer. He identified document.oai.annotation.request(). The new development is a concrete integration experiment following his earlier tool introduction; it does not demonstrate automatic messaging.
Official documentation limits the API to supported built-in browser contexts on secure top-level pages. A user action triggers it and permission may be needed. Opening an annotation sends no message; the user submits it. An accepted result does not prove that the editor opened.
Baoyu also says Cursor and Claude Code lack this API, within his analysis. We did not reproduce the experiment or independently inspect every interface in those products. For developers, the practical lesson is to check public integration and submission boundaries before choosing reverse engineering.
Zhou Xianwang gets 14 years at first instance; this is a bribery case
BBC News 中文 · @bbcchinese
BBC Chinese reported on October 10 that Shangqiu’s intermediate court sentenced former Wuhan mayor Zhou Xianwang to 14 years for bribery, imposed a RMB5 million fine and ordered recovery of proceeds. Citing the court, it gives more than RMB60.5 million in bribes between early 2003 and May 2025.
The reported conduct involved using public offices to help with contracts, business operations and administrative approvals. Mitigation included attempted offenses, truthful disclosure, admitting additional conduct, remorse and restitution. This is a first-instance sentence; we have no subsequent appeal or final-judgment evidence.
Zhou’s 2020 remarks about needing authorization to disclose outbreak information are historical context. The BBC says the court notice did not confirm a connection to Jiang Chaoliang’s case. The evidence does not establish a conviction for the controversy over pandemic disclosure.
This catches up an October 10 report, not a new judgment today. We read the full BBC article; the court’s original WeChat notice was inaccessible. Figures and sentencing grounds remain attributed to the BBC’s account, not independently verified against the court text.