Daily news: Oct 22, 2026

Tesla VP Ashok Elluswamy said during the company’s third-quarter earnings call that owners will be able to add their own vehicles to Tesla’s robotaxi fleet in the second quarter of 2027. The article notes this follows years of Elon Musk promising unsupervised autonomous driving and owner participation in a Tesla robotaxi network.

Blackstone’s Jas Khaira said the firm is financing AI infrastructure through BXN1, including a $5B commitment to Crux AI, backing chip startup Etched, and debt structures such as CoreWeave’s early $2.3B loan and Firmus’ $10B credit facility. Concerns have grown as CoreWeave’s debt reached about $50B, its shares fell more than 20%, and Firmus canceled a planned IPO at roughly a $30B valuation, but Khaira argued much data center debt is ultimately tied to highly creditworthy hyperscaler customers. Blackstone is also pushing AI adoption through Ode, a $1.5B joint venture with Anthropic and Hellman & Friedman, though Khaira said ROI across portfolio companies is “uneven.”

Microsoft has rolled out a la carte Xbox Cloud Playtime purchases to Xbox Insiders, letting non-subscribers stream eligible games they already own on supported devices. Introductory pricing is $4 for 10 hours, $8 for 25 hours, or $12 for 50 hours, and purchased time remains available until used.

Railway detected anomalous activity on Oct. 21, 2026 involving systems that store customer environment variables, saying an unauthorized third party may have accessed encrypted variables and decrypted contents for some of them. The company recommends all users immediately rotate API keys, environment variables, and private registry credentials, and says deployments are paused until builds are restored.

NVIDIA introduced Kumo Relational, a pre-trained relational foundation model that makes predictions directly on enterprise databases from Predictive Query Language queries, using up to 10,000 historical examples as context in a single forward pass. NVIDIA says it needs no task-specific training, feature engineering, or table flattening, and reports that it beat supervised relational deep learning and engineered gradient-boosted tree pipelines across 41 tasks on 15 databases, including an 8% MRR gain over AutoGluon ensembles on SAP SALT. It is offered through NVIDIA's open-source GPU-native structured-data-models library or as an NVIDIA NIM, with claimed scaling to databases above 500B rows.

GitHub says 99.99999% of repositories now live in both its data centers and Azure, with most repos using two of five replicas in Azure and Azure serving about half of Git traffic. The remaining busiest repositories are constrained by a 25ms data-center-to-Azure round trip, so GitHub is shifting from repository-wide write serialization to per-reference serialization, which showed 3–4x higher throughput at lower latency in internal tests. GitHub also outlines planned virtual replicas backed by Azure Blob Storage to separate read scaling from write quorum behavior.

The article says a rush of new and proposed data centers in Texas, driven by generative AI infrastructure, could raise air pollution toward levels last seen in the 2000s. The impact depends on whether developers and utilities meet the huge electricity demand with gas turbines or renewable energy, and the excerpt says current signs look unfavorable.

Meta confirmed plans to create a publicly traded class of non-voting stock that could be used for capital raising, employee compensation, acquisitions, and other corporate purposes while preserving existing Class A and B voting power. The move comes as Meta reported $64 billion in quarterly revenue, up 26% year over year, while costs and expenses rose 68%, and CEO Mark Zuckerberg said more than 10 million people in the U.S. and Canada have set up Muse personal agents in about six weeks.

A survey commissioned by Origina found that enterprises with at least 10,000 employees spend an average of 28% of software budgets on unwanted upgrades, migrations, and licensing changes. The survey covered 2,018 CIOs and senior IT decision-makers at companies with at least $1B in annual revenue in the US and UK, but Origina sells independent software maintenance, which may bias the framing.

Dharma-AI previews VocabX, a vocabulary-specialization method tested on a production Portuguese essay grader built on Qwen3-4B. By weighting token choices by measured serving cost rather than frequency, it reduced the output vocabulary from about 152,000 to 65,000 rows and cut measured inference cost by 77% without changing the base model, data, or training recipe.