According to Lex Fridman on X, he recorded a long-form, technical deep-dive podcast with NVIDIA CEO Jensen Huang and plans to release it on Monday, highlighting NVIDIA’s role as the world’s most valuable company by market cap and the engine powering the AI revolution (source: Lex Fridman on X). As reported by Lex Fridman, the conversation focused on on- and off-mic technical topics, signaling insights likely to cover GPU roadmaps, data center-scale AI infrastructure, and model training efficiency that directly impact AI compute supply chains and total cost of ownership (source: Lex Fridman on X). For businesses, the expected discussion points imply near-term opportunities in optimizing inference with next-gen NVIDIA platforms, expanding AI cloud partnerships, and refining MLOps around accelerated computing to capture demand in generative AI and enterprise LLM deployment (source: Lex Fridman on X). (Source)

According to Ethan Mollick on X, his experiment used ChatGPT 5.4 Pro to estimate how “lucky” a person is to live today by benchmarking historical lifestyles against a modern middle-class baseline, finding that only about 1.5% of the roughly 117 billion humans who ever lived matched or exceeded a contemporary middle-income lifestyle; as reported by Ethan Mollick, this showcases a concrete use of large language models for data synthesis, scenario framing, and public communication of quantitative history. According to Ethan Mollick, framing the analysis as a time traveler's veil of ignorance illustrates how LLMs can structure counterfactuals, normalize metrics across eras, and communicate results for policymaking and education. As reported by Ethan Mollick, such LLM-powered historical benchmarking creates opportunities for AI consultancies to build reproducible pipelines for long-horizon economic comparisons, develop explainable prompts and toolchains for data validation, and offer decision-support products for think tanks and foundations evaluating progress and welfare over time. (Source)

According to Ethan Mollick on Twitter, large language models lag in writing because they lack an objective judge and exhibit poor subjective self-judgment, limiting self-improvement. As reported by Christoph Heilig’s blog, experiments show GPT‑5.x can be steered by pseudo‑literature prompts to overrate weak prose, revealing evaluation misalignment and vulnerability to style hacks (source: Christoph Heilig). According to Heilig, these failures undermine reward-model reliability and RLHF pipelines that depend on model or human preferences for literary quality, constraining progress in long-form generation. For businesses building AI writing tools, the cited evidence implies opportunities in external objective metrics, multi-rater human annotation markets, and retrieval-augmented critique systems to stabilize quality judgments and reduce reward hacking (source: Christoph Heilig). (Source)

According to Fox News AI on Twitter, investors are funding AI-powered cow collars as U.S. cattle herds decline and beef prices rise, aiming to boost productivity and animal health through precision livestock monitoring (source: Fox News). As reported by Fox News, these smart collars use sensors and machine learning to track grazing, heat detection, location, and early illness markers, helping ranchers reduce feed costs and improve calving rates. According to Fox News, the business case hinges on higher carcass weights and lower mortality, offering faster payback amid elevated beef prices. As reported by Fox News, adoption is accelerating as ranchers seek real-time insights, automated alerts, and integration with herd management software, creating opportunities for hardware makers, data platforms, and agtech financiers. (Source)

According to pictoryai on X, the Pictory PowerPoint Add-in converts PPT slides into export-ready videos directly inside PowerPoint, streamlining script-to-video workflows for marketing and training teams. As reported by Pictory Academy, users can install the add-in from Microsoft AppSource, select a template, auto-sync slide text to voiceover, and export MP4 with captions, enabling faster content repurposing and brand-consistent video creation at scale. According to Pictory Academy, the add-in automates narration, scene timing, stock footage insertion, and subtitle generation, reducing manual editing time and lowering video production costs for SMBs and enterprises focused on sales enablement, e-learning, and social snippets. As reported by Pictory Academy, this positions AI-powered video inside familiar Office workflows, creating opportunities for agencies and internal comms teams to batch-convert existing decks into bite-size clips for LinkedIn, YouTube, and LMS libraries. (Source)

According to Microsoft Copilot on X, creator Paul Couvert (@itsPaulAi) demonstrated Copilot Tasks completing an end-to-end workflow in a single prompt, including using a cloud browser to choose tools, interacting with webpages to input data, generating a PowerPoint, and drafting an email via Outlook, with real-time access to Word, PowerPoint, Excel, and Outlook (source: Microsoft Copilot post citing @itsPaulAi video on X). As reported by the same post, Copilot Tasks also supports scheduling recurring automations, enabling weekly or monthly runs without manual intervention. For businesses, this indicates practical opportunities to automate research-to-report pipelines, sales collateral creation, and meeting follow-ups directly within Microsoft 365, reducing context switching and accelerating throughput. (Source)

According to Fox News AI on Twitter, Amazon Health AI promises a pocket doctor experience via a new suite of healthcare-focused generative AI tools integrated with Alexa and Amazon Clinic, aiming to streamline symptom triage and care navigation (as reported by Fox News). According to Fox News, the service leverages medical question answering and automated visit summaries to reduce clinician documentation time and improve patient intake conversion in virtual care workflows. As reported by Fox News, Amazon positions the platform for payers, providers, and telehealth startups by offering APIs for compliant data handling and EHR integration, highlighting opportunities to cut contact center costs and boost patient self-service. According to Fox News, the initiative underscores a competitive push against Google and Microsoft in healthcare AI, with business upside in white-labeled triage bots, remote monitoring support, and employer health benefits tools. (Source)

According to Greg Brockman on X, the latest Codex hackathon showcased over 200 projects with the Top 5 featuring advanced multi‑agent coding orchestration across different providers and C++ firmware for brainwave readers, demonstrating rapid prototyping potential for autonomous developer tools and human‑computer interfaces (source: Greg Brockman citing Gabriel Chua). As reported by Gabriel Chua on X, one team ran Codex agents continuously while exploring Ho Chi Minh City, indicating robust hands‑off reliability for background code generation workflows, which could lower engineering costs for startups and accelerate continuous integration pipelines. According to the organizers LotusHack, GenAI Fund, and HackHarvard credited in the thread, the event underscores growing demand for cross‑provider agent orchestration stacks, creating business opportunities for tooling vendors in agent routing, evaluation, and observability. (Source)

According to Microsoft Copilot on X, creator Paul Couvert (@itsPaulAi) demonstrated Copilot Tasks completing an end‑to‑end workflow from a single prompt: using a cloud browser to identify a tool, interacting with a webpage to input data, interpreting on-page information, generating a PowerPoint, and drafting an Outlook email, with optional scheduling for recurring runs (source: Microsoft Copilot post citing @itsPaulAi video on X, Mar 22, 2026). As reported by the same post, Copilot Tasks operates across Word, PowerPoint, Excel, and Outlook in real time, positioning it as a strong alternative to Anthropic’s Claude Cowork for task automation. According to @itsPaulAi’s video, the seamless first-try execution highlights business opportunities in automated report generation, sales outreach cadences, and monthly KPI packs, particularly for Microsoft 365 tenants seeking integrated agentic workflows. (Source)

According to Ethan Mollick on X, a preregistered study of 968 participants found almost no correlation between feeling empathic and communicating empathy, but a single practice session with an AI coach measurably improved empathy communication skills. As reported by the study authors on arXiv, participants who practiced responses with an AI coach showed statistically significant gains on validated empathy communication scales after one session, indicating rapid skill transfer in hard-to-teach social behaviors. According to the arXiv paper, the intervention used structured practice with feedback to close the gap between internal empathic concern and externally observable empathic communication, highlighting immediate applications for customer support, healthcare triage, and manager training programs. As reported by arXiv, the preregistered design and large sample size strengthen external validity for enterprise learning and development teams evaluating AI-enabled soft-skills training. (Source)

According to Sawyer Merritt on X, Tesla is testing Robotaxi service in Phoenix, Arizona using a Model Y equipped with rear camera washers and a California manufacturer plate, aligning with Tesla’s Q4 earnings call guidance that Phoenix is among seven metro areas targeted for robotaxi coverage in H1 2026; according to Tesla’s Q4 2025 earnings call remarks, this pilot signals progress toward supervised commercial robotaxi operations, with enterprise opportunities in autonomous ride-hailing, fleet optimization, and data-driven safety validation in the Phoenix market. (Source)

According to God of Prompt on Twitter, a premium AI bundle offers best-in-class marketing and business prompts, unlimited custom prompt creation, n8n workflow automations, and weekly updates with lifetime access at godofprompt.ai/pricing. As reported by the original post, the package centralizes prompt engineering assets that can accelerate campaign copy, ad variations, and CRM outreach while n8n automations can orchestrate lead capture, enrichment, and content generation pipelines end to end. According to the same source, weekly updates indicate ongoing prompt library expansion, which can reduce prompt maintenance costs and improve model output quality over time. For businesses, this creates opportunities to standardize prompt operations, shorten go-to-market cycles, and scale content production across GPT-class and Claude-class models while leveraging n8n to integrate LLM calls with marketing stacks such as email, CRMs, and analytics. (Source)

According to God of Prompt on Twitter, a new AI research paper is available at arXiv under identifier 2603.18908. As reported by arXiv via the linked abstract page, the paper is publicly posted at https://arxiv.org/abs/2603.18908, but no additional metadata was provided in the tweet to detail the model, method, or benchmarks. According to arXiv, accessing the abstract and PDF will provide verified details on the proposed technique, datasets, and results, which are essential for assessing business impact such as model performance gains, compute requirements, and deployment feasibility. For AI product teams and investors, the immediate opportunity is to review the arXiv abstract and methods section to identify potential commercialization paths, licensing constraints, and integration points with existing MLOps stacks, according to standard arXiv usage and citation practices. (Source)

According to God of Prompt on X, citing a new Columbia University paper, independent frontier models like GPT, Gemini, Qwen, Mistral, and Cohere exhibit high cross-model CKA similarity (0.595–0.881), enabling a single affine map to align internal representations for private inference (as reported by the Columbia study via the X thread). According to the thread, the HELIX system replaces full-transformer encrypted inference—previously 25–281GB per query and 20–60s latency—with a linear alignment plus homomorphic encrypted classification, achieving sub-second latency and under 1MB communication with 128-bit CKKS security. As reported by the same source, HELIX trains the alignment map using encrypted client embeddings on public data, then runs inference by locally applying the alignment, encrypting the transformed features, and letting the provider perform a single linear operation; the provider never sees plaintext inputs or model weights. According to the X post, tokenizer compatibility strongly predicts cross-model generation quality (r=0.898), and models over 4B parameters with tokenizer match rate above 0.7 can generate coherent text across families using only a linear transform. Business impact: according to the Columbia results as relayed by God of Prompt, enterprises in regulated sectors could cut private LLM inference costs and latency by orders of magnitude, unlocking viable deployments for hospitals, banks, and legal firms that cannot share raw data with third-party providers. (Source)

According to Greg Brockman on X, subagents in Codex are very powerful. As reported by his post, the highlight is Codex’s ability to coordinate specialized subagents for tasks like code generation, refactoring, and tool use, enabling parallel problem decomposition and faster turnaround for complex software tasks. According to OpenAI documentation referenced by developers, multi-agent patterns can improve success rates for long-horizon coding by delegating linting, testing, and API integration to focused workers under a supervisor agent. For businesses, this suggests new product opportunities in autonomous code assistants, CI automation, and enterprise integration pipelines that capitalize on subagent orchestration and tool calling. (Source)

According to Ethan Mollick on X, Claude with Computer Use autonomously downloaded the NetHack codebase, read documentation, and began implementing a new horror-inspired creature by modifying source files until hitting rate limits, demonstrating concrete agentic capabilities for software development workflows (as reported by Ethan Mollick’s X post and thread). According to Mollick’s post, the model executed multi-step tool use including repository fetch, file inspection, and targeted code edits, highlighting near-term applications in rapid prototyping and legacy code maintenance for game development and enterprise software. As reported by Ethan Mollick, the run-by-run trace suggests viable business use cases such as automated feature insertion, refactoring, and test generation under human supervision, with constraints around API rate limits and oversight requirements. (Source)

According to Ethan Mollick on X (Twitter), OpenAI's Codex autonomously downloaded NetHack, modified game items to increase player power, and produced a working Windows .exe, overcoming environment and build issues that previously stymied older AI tools. As reported by Mollick’s post, this showcases practical code synthesis, dependency management, and build orchestration—key capabilities for AI software agents. For businesses, this indicates near-term opportunities to automate legacy app refactors, rapid prototyping, and modding workflows; according to Mollick, the successful artifact delivery (.exe) is evidence of reliable multi-step tool use that can reduce developer cycle time and QA overhead in controlled pipelines. (Source)

According to SawyerMerritt on X, Tesla's Dojo D3 chip is being used inside SpaceX AI satellites, with a posted image and link suggesting on-orbit inference hardware integration; however, independent confirmation is not provided in the post. As reported by the X post, the claim implies edge AI processing in space for tasks like onboard vision, autonomy, and RF signal classification, reducing ground downlink needs and latency. According to prior Tesla disclosures referenced by industry coverage, Dojo is designed for high-throughput training, and if a D3 variant is space-hardened for inference, it signals a vertical stack from Tesla silicon to SpaceX satellite operations, potentially lowering cost per inference and enabling real-time services. As reported by the post, if validated by SpaceX or Tesla, business opportunities include satellite-based AI analytics, premium enterprise APIs for geospatial intelligence, and cross-division silicon monetization. (Source)

According to Sawyer Merritt on Twitter, SpaceX released a new video of a lunar electromagnetic mass driver to launch large AI satellites using Tesla chips; however, no corroborating report or official release from SpaceX, Tesla, or reputable outlets confirms this claim as of now. According to SpaceX’s official channels and newsroom, there is no press release or technical brief on a Moon-based mass driver or AI satellites powered by Tesla silicon. As reported by Tesla’s investor relations and product pages, Tesla develops FSD and Dojo chips for automotive and data center use, but no source confirms their deployment in SpaceX satellites. Given the lack of verification, businesses should treat this as unconfirmed and avoid operational decisions until an official statement appears from SpaceX or Tesla. (Source)

According to Sawyer Merritt on X, Elon Musk said the cost of deploying AI in space will fall below the cost of terrestrial AI within 2–3 years, noting that operations in space get easier over time. As reported by Sawyer Merritt, this implies near-term opportunities for space-based inference at scale—such as Earth observation analytics, inter-satellite routing, and edge model serving on Starlink-class constellations—where reduced thermal constraints and abundant solar power could lower total cost of ownership versus ground data centers. According to the cited post, if realized, companies building radiation-hardened accelerators, on-orbit model update pipelines, and space-to-cloud MLOps could gain first-mover advantages in latency-sensitive markets including disaster monitoring, maritime tracking, and global connectivity. (Source)