In the early 20th century, mathematician Kurt Gödel shattered the hope that mathematics could be both complete and consistent. His incompleteness theorems rely on a clever encoding: assign a unique natural number to every logical statement by mapping symbols to primes and exponents. This lets arithmetic talk about itself. Gödel then constructed a statement G that says 'G cannot be proved.' If G is false, its negation is provable, leading to contradiction; if true, it's unprovable, showing the system's incompleteness. Scientific American's Manon Bischoff walks through the encoding step by step, from assigning numbers to basic operations to building self-referential sentences. The result is a landmark in logic: any formal system rich enough to contain arithmetic will always contain true but unprovable statements. The article makes the construction concrete, showing how prime factorization ensures unique decoding, and why this discovery still shapes our understanding of mathematical truth.
Gödel published his incompleteness theorems in 1931, upending David Hilbert's program to axiomatize all of mathematics. The theorems show that in any consistent formal system capable of basic arithmetic, there exist statements that can neither be proved nor disproved within the system. Gödel numbering is the technique that makes this self-reference possible.
Microsoft announced its Majorana 2 quantum chip this week, claiming improved stability for its topological qubits. The company says the new chip, which uses lead-based superconductors, extends qubit lifetime to 20–60 seconds and could enable scalable quantum computing by 2029. Outside physicists are not convinced. Henry Legg of the University of St. Andrews notes that the preprint shows data from only a handful of instances on a single device, far short of the reproducibility standard the field requires. Microsoft has a history of retracted claims in this area: a 2021 Nature paper was pulled after experts found the data could be explained by material defects rather than Majorana quasiparticles. The new preprint has not been peer-reviewed, and critics say it would not pass review if submitted by any other group. The episode highlights the gap between corporate press briefings and the evidence bar in condensed matter physics.
Topological quantum computing aims to use quasiparticles called Majorana zero modes to create qubits that are inherently resistant to noise. Microsoft has pursued this approach for years, but previous announcements have been met with skepticism after retractions and disputed data. The Majorana 1 chip, announced in 2025, faced similar criticism.
President Trump has named Bill Pulte, the director of the Federal Housing Finance Agency, as acting Director of National Intelligence, replacing Tulsi Gabbard when she steps down on June 30. Pulte has no intelligence background. He comes from a home-building dynasty and private equity, and has used his housing agency role to make criminal referrals against Trump's political opponents, including Senator Adam Schiff and New York Attorney General Letitia James. None of those referrals led to prosecutions, and the Government Accountability Office is now investigating the FHFA's mortgage fraud referral process. Democrats condemned the pick; Senator Mark Warner called it evidence that Trump expects the intelligence chief to pursue political retribution. Even some Republicans expressed unease: Senator John Cornyn said he saw no evidence of Pulte's qualifications. Pulte can serve up to 210 days without Senate confirmation.
The Director of National Intelligence oversees 18 US intelligence agencies. The position requires Senate confirmation for a permanent appointee, but acting officials can serve temporarily. Tulsi Gabbard announced her resignation effective June 30. Pulte will simultaneously remain FHFA director and chairman of Fannie Mae and Freddie Mac.
President Trump signed an executive order Tuesday creating a voluntary framework for AI companies to share frontier models with the federal government before public release. Companies can submit models up to 30 days in advance and receive confidentiality protections, but participation is not mandatory. The order directs agencies to develop methods for assessing AI models' cybersecurity risks, especially for critical infrastructure. It explicitly states it is not a licensing or preclearance regime. The move follows Anthropic's limited April rollout of its Mythos model, which flagged thousands of high-severity vulnerabilities in major operating systems and browsers. Google, Microsoft, and xAI had already agreed to pre-release review by the Commerce Department's Center for AI Standards and Innovation. The order has drawn praise from groups like Americans for Responsible Innovation, who urged Congress to codify mandatory protections.
The Biden administration had secured voluntary AI safety commitments from OpenAI and Anthropic in 2024. Trump initially took a hands-off approach under former AI czar David Sacks, but the Mythos model's capabilities and the broader AI safety debate have shifted the administration's stance toward some oversight.
Florida's Republican-controlled legislature has passed a bill giving statewide education boards the power to amend general education requirements at public colleges and universities. The Board of Governors and Board of Education already removed sociology from the list of approved gen ed courses this spring, with officials calling the discipline 'social and political advocacy dressed in the regalia of the academy.' The new legislation, House Bill 5601E, would let those boards not just approve or reject course lists but rewrite them. The bill now awaits Governor Ron DeSantis's signature. Faculty union president Robert Cassanello predicted 'wholesale amending' of gen ed statewide. Critics, including Democratic lawmakers and PEN America, argue the move erodes academic freedom and places curriculum decisions in the hands of political appointees rather than educators. The bill also transfers a University of South Florida campus to New College of Florida, which has been reshaped by conservative trustees.
Florida has become a focal point for conservative higher education reform under Governor DeSantis, with previous actions including the takeover of New College's board and restrictions on diversity programs. The sociology course removal in March 2026 marked the first time statewide boards directly intervened in curriculum content, citing ideological concerns.
Private nonprofit colleges are discounting tuition at record rates, according to new data from the National Association of College and University Business Officers. The average discount rate for first-time undergraduates is projected to hit 57.1% in the 2025-26 academic year, up from 54.5% the year before. For all undergraduates, the rate is expected to reach 51.3%. About 90% of first-year students now receive some institutional aid. Despite the deeper discounts, net tuition revenue per student fell 1.9% after inflation in 2024-25, and revenue from first-time students dropped 2.2%. NACUBO noted that retention alone is not enough to offset financial strain at tuition-dependent institutions. The study, based on a survey of 258 private nonprofits, underscores the persistent challenge of a shrinking pool of high school graduates and the pressure to offer larger scholarships to fill seats.
The tuition discount rate measures institutional grant aid as a percentage of gross tuition revenue. A rising discount rate means colleges are giving back a larger share of sticker-price tuition in the form of scholarships, often funded by tuition revenue from full-pay students or other sources. The trend has accelerated as the number of high school graduates declines.
A study led by Stanford Law professor Julian Nyarko found that law professors overwhelmingly prefer AI-generated answers to student questions over responses written by their peers. In a blind evaluation of nearly 3,000 comparisons across 16 law professors, AI answers won 75% of head-to-head matchups. The professors rated AI responses as pedagogically harmful only 3.5% of the time, compared to 12% for human-written answers. The study focused on contract law, a domain requiring judgment and nuanced reasoning rather than factual recall. Researchers calibrated AI responses to match human answer length and structure, and used multiple evaluation methods. Co-author Sarath Sanga of Yale Law noted that law often lacks clear right answers, making the result more striking. The findings suggest AI tutors could provide high-quality, on-demand support in judgment-rich fields, potentially broadening access to expert guidance in legal education.
The study, titled 'Law Professors Prefer AI Over Peer Answers,' was conducted by researchers from Stanford, Yale, NYU, and the University of Chicago. It tested large language models as tutors for contract law courses. Previous AI evaluations have mostly focused on subjects with clear correct answers, making this study notable for its application to a field requiring argumentation and ambiguity.
Google has been quietly offering to pay Android app developers for access to their codebases to train AI coding tools, 404 Media reports. An email sent to a developer with millions of downloads describes a 'confidential content offer pilot' that would license code from active apps and archived projects on a non-exclusive basis, with developers retaining intellectual property rights. The email does not mention AI, but a linked page frames the program as a partnership to improve Google's AI products. The move suggests Google is struggling to match competitors like Anthropic and Microsoft in AI code generation and is seeking high-quality, real-world code beyond what it can scrape from the web. Google previously paid Reddit $60 million for training data. The developer who shared the email requested anonymity, fearing retaliation for disclosing details of the confidential program.
AI code generation tools like Anthropic's Claude Code and Microsoft's GitHub Copilot have gained significant traction. Training these models requires large volumes of high-quality code. While public repositories provide some data, proprietary, real-world codebases offer examples of production-level patterns, error handling, and architecture that are underrepresented in open-source code.
Labor economist Kathryn Anne Edwards rejects the Silicon Valley vision of a permanently unemployed 'idle class' created by AI. In an interview with Platformer, she argues that American workers are resilient and adaptable, and that writing them off as unemployable is classist. But she also warns that the US safety net is broken and unprepared for any mass displacement, whether from AI or other shocks. Edwards points to concrete fixes: overhauling unemployment insurance, decoupling health insurance from employment, subsidizing relocation for job seekers, and raising the estate tax. She calls herself a cynical optimist: the fact that the US has never seriously tried to build a robust safety net means there is enormous room for improvement. The conversation is part of Platformer's miniseries on AI and jobs, which has featured tech leaders who mostly downplay disruption.
Kathryn Anne Edwards is a labor economist and independent policy consultant who writes for Bloomberg Opinion and co-hosts the Optimist Economy podcast. The interview follows earlier episodes with Box CEO Aaron Levie, Google's James Manyika, and Claude Code creator Boris Cherny, all of whom argued that AI will change jobs more than eliminate them.
Lumafield, an industrial CT scanning company, has published detailed scans of BYD car components, revealing the engineering behind the Chinese automaker's vertical integration. The analysis examines a lithium iron phosphate (LFP) battery cell, showing two parallel jellyroll electrode stacks, consistent layer alignment, and a pressure relief valve. The scans also capture manufacturing imperfections: uneven winding tension in the electrode rolls that could accelerate degradation over the cell's lifetime. A second scan dissects the driver door control panel from a BYD Tang SUV, which consolidates mirror, window, and lock controls into a single networked module. The PCB layout maps directly to a LIN bus network, illustrating how BYD reduces subassembly count by designing both hardware and firmware in-house. The article provides a rare, non-destructive look inside mass-market EV components and highlights the tradeoffs in BYD's cost-focused engineering approach.
BYD (比亚迪) is a Chinese automaker that designs and manufactures its own batteries, unlike most Western competitors who source cells from suppliers like Panasonic or LG. Its Blade battery uses LFP chemistry, which sacrifices some energy density for lower cost, better thermal stability, and longer cycle life. The company's vertical integration extends to electronics and software.
