Skip to content
Christmas & Holiday Lights 2026Explore the community map

Frontier Engineering & Enterprise AI

OpenAI Scales Hardware AI as DeepMind and Palantir Post Gains

OpenAI partners with Mazda and Chip Ganassi Racing on GPT-6 Astra engineering while Google DeepMind launches Cheminformatics Skills and Palantir revenue rises 96% to $2.31 billion.

By East Bay Post · Syndicated from Lazy Tensor
Syndicated reporting from Lazy Tensor · Lazy Tensor
Published · 6 min read

Editorial print illustration on warm cream paper showing a precision mechanical caliper, drafting compass, and molecular ring lattice on an engineering drafting table
Frontier AI laboratories and enterprise software providers expanded into physical CAD engineering, computational chemistry, and commercial agent deployments on Tuesday. · Lazy Tensor / Editorial Visual Desk
Your listening Queue & catch up
Unheard stories first, timed at 1×. Your selections stay first.

Add a story or build a catch-up queue to start listening.

SAN FRANCISCO — Frontier model developers and enterprise software companies announced a wave of physical engineering, computational science, and consumer intelligence milestones on Tuesday, November 3, led by OpenAI expanding hardware design workflows with GPT-6 Astra, Google DeepMind introducing agentic cheminformatics on Google Antigravity, and Palantir reporting $2.31 billion in third-quarter revenue.

OpenAI Scales GPT-6 Astra for Physical Engineering with Chip Ganassi Racing and Mazda

According to OpenAI News, OpenAI’s internal hardware and infrastructure teams—which use Codex and ChatGPT to generate and revise computer-aided design (CAD) models, prepare manufacturing drawings, and configure simulations across robotics and compute infrastructure—have increased their model usage measured in tokens 23-fold since February.

To test end-to-end multimodal and computer-use engineering, 2 OpenAI team members used GPT-6 and GPT-6 Astra to define functional requirements, design a mechanical enclosure (drawing on a Work Louder collaboration), place components on a printed circuit board (PCB), and write firmware for a hardware device that lets users talk to Codex. The design took under 1 week as a side project, and a contract manufacturer built a small batch of prototypes in under 1 month. On BenchCAD 2.0, which evaluates 3D part reconstruction and circuit board connection mapping, a newer internal OpenAI model achieved an average score of 0.875, compared with 0.594 for GPT-6 Astra at its highest reasoning setting.

OpenAI also detailed 2 industrial engineering collaborations:

  • Chip Ganassi Racing (CGR): Working with CGR’s research and development team to convert 3D scans of existing racing components into editable CAD models and 2D manufacturing drawings when original design files are unavailable.
  • Mazda Motor Corporation: Collaborating with Mazda’s aerodynamics and crash-safety engineers alongside OpenAI’s Forward Deployed Engineering teams to model new vehicle nose shapes, simulate airflow and structural behavior, and reduce vehicle drag.

Google DeepMind Launches Cheminformatics Skills on Google Antigravity

In a technical report authored by Zoi Tsangalidou, Ruijie (Kelly) Zhang, Sukhdeep Singh, and Michael V. LeVine, Google DeepMind introduced a suite of Cheminformatics Skills for Google Antigravity, grounding agentic scientific reasoning in deterministic chemistry calculations via the open-source RDKit toolkit and connectors to ChEMBL and PubChem.

Focusing on non-small cell lung carcinoma tyrosine kinase inhibitors (TKIs), the team demonstrated how medicinal chemists rigidify molecules to eliminate the entropic penalty of rotatable bonds—comparing first-generation TKI crizotinib, which has 5 major rotatable bonds, with third-generation macrocycle lorlatinib, which is cyclized to reduce effective rotatable bonds to 0. In 1 hour, an Antigravity agent searched thousands of literature compounds, filtered out buffers and solvents, launched an interactive scatter-plot viewer, executed matched molecular pair analysis (MMPA), benchmarked potency prediction models across 4 experimental readouts using chronological and scaffold splits, and proposed novel rigidified macrocycles complete with curved-arrow electron-pushing diagrams and synthetic routes ready for AlphaFold or molecular dynamics evaluation.

To accelerate adoption, Google is offering 10,000 scientists the highest usage tier in Google Antigravity free of charge for 6 months on a rolling basis.

Palantir Q3 Revenue Surges 96% to $2.31 Billion as U.S. Commercial Tops Government

As reported by TechCrunch, Palantir reported that third-quarter revenue rose 96% year-over-year—and 20% from the prior quarter—to $2.31 billion, surpassing the $2 billion mark for the first time and beating Wall Street consensus expectations of $2.18 billion. Adjusted earnings per share reached $0.44, ahead of analysts’ $0.41 forecast.

For the first time, Palantir’s U.S. commercial revenue exceeded its U.S. government revenue, with $984 million in U.S. commercial contracts (up 148% year-over-year) compared with $953 million in U.S. government revenue. However, total contract value (TCV) from U.S. enterprises declined to $1.46 billion from a record $2.13 billion in the previous quarter as corporate buyers opted for shorter-duration initial trials. Chief Executive Officer Alex Karp expressed confidence in contract expansions and raised full-year U.S. commercial revenue guidance to in excess of $3.559 billion, representing a growth rate of at least 143%.

ChatGPT Reaches 20% of Google Search Volume; Jill Hazelbaker Joins OpenAI

At a Monday press event reported by The Verge, Adam Fry, OpenAI’s consumer product lead for ChatGPT, announced that the volume of web-search queries run through ChatGPT now equals 20% (20 percent) of the search volume run through Google—compared with the roughly 3% (3 percent) search market share Microsoft CEO Satya Nadella cited for Bing during Google’s 2023 antitrust trial. Fry also highlighted OpenAI’s visual "Intelligent UI" canvas responses and plans to introduce sponsored product images in ChatGPT.

Separately, OpenAI News announced that Jill Hazelbaker will join OpenAI as Chief Corporate Affairs Officer to lead global communications and corporate positioning. Hazelbaker previously served for a decade at Uber as President and Chief Corporate Affairs Officer, led communications and public policy at Snap, held senior communications and European government relations roles at Google, and served as National Communications Director for Senator John McCain’s 2008 presidential campaign and Press Secretary to New York City Mayor Michael Bloomberg.

Apple Rolls Out iOS 27.3 Siri App Skills and watchOS 27.3 Audio Intelligence; Netflix Debuts AI Discover

Consumer platforms also shipped major AI updates on Tuesday:

  • Apple iOS 27.3 & watchOS 27.3 Developer Betas: As reported by The Verge, Apple’s iOS 27.3 Developer Preview allows Siri AI to perform actions across third-party apps using on-screen awareness, personal context, and a new "app skills" feature during Live Activities, alongside expanded capabilities across 15 native Apple apps (Calendar, Files, Finder, Journal, Mail, Maps, Messages, Music, Notes, Photos, Preview, Reminders, Safari, Shortcuts, and Weather). In a companion update reported by The Verge, the watchOS 27.3 developer beta introduces Audio Intelligence on Apple Watch—featuring Siri Recap conversation summaries and Live Rewind, which displays a text transcript of the previous 15 seconds of audio when a user double-presses the Digital Crown—alongside a redesigned Health app in iOS 27.3 with an Insights tab and expanded support across 11 language groups (global English, Simplified and Traditional Chinese, French, Japanese, Korean, Norwegian, Portuguese, Spanish, Turkish, and Vietnamese).
  • Netflix AI Discover Tab: According to TechCrunch and The Verge, Netflix is rolling out a new "AI Discover" tab (marked by a red gem icon) on supported TVs for English-speaking U.S. members following recent layoffs of around 1,000 employees, surfacing mood-based themes such as "Something Light," "Family Movie Night," "Quick Watches," and "scare me" with real-time "Tune Your Recommendations" feedback controls.

Sources

Comments (0)

No comments yet. Be the first to share a thought on this story.