AI News

In a wide-ranging interview with Vox’s Future Perfect, behavioral economics co-founder George Loewenstein discusses his new book *The Opposite of Happiness: How Bad Feelings Make and Break Us*, arguing that misery is not a malfunction but „the system” itself—an evolutionary tool that motivates survival. Loewenstein explains how negative emotions like hunger, pain, and loneliness are more powerful and persistent than positive ones, and how modern phenomena like social media exploit our evolved programming for profit rather than our benefit. He also introduces the concept of „meta-emotions”—the guilt we feel about feeling bad—suggesting that our cultural obsession with happiness compounds our suffering by making us ashamed of ordinary negative feelings. Ultimately, Loewenstein advocates for insight over self-help and argues that reducing misery, through policies like robust safety nets, should be a central goal of public policy.

Mistral AI has launched Mistral Large 4, a new flagship open-weight model now available in public preview ahead of its October 27 rollout. The model processes both text and images and delivers advanced agentic, cybersecurity, geospatial, and grounding capabilities, reaching state-of-the-art performance on benchmarks like SWE-bench and Terminal Bench. Trained on more than 160 languages, Large 4 is positioned among the world’s best open-weight models, closing the gap with Chinese competitors that have dominated the sector. Mistral says the preview period will gather feedback from developers, cybersecurity experts, and state officials to support further reinforcement learning and fine-tuning.

Companies of all sizes are increasingly turning to a DIY approach to software, using generative AI tools like ChatGPT and Codex to build internal applications instead of purchasing costly add-on features from enterprise vendors such as Workday, SAP, and Salesforce. A recent McKinsey survey of 1,719 executives found that about a third of organizations decided against buying at least one software product or feature because they could build it themselves with AI. While most companies aren’t replacing entire enterprise platforms, this trend could erode the incremental revenue vendors earn from add-on features, with analysts warning that „the vendor’s most dangerous competitor may now be the customer.” Despite risks around maintenance, compliance, and accuracy, firms like West Monroe, Spotify, Twilio, and Gray are realizing significant savings—Gray alone expects to avoid $1 million in vendor and consulting fees by building a financial forecasting tool in-house.

In 2026, venture capital has concentrated at the top, with AI startups capturing 61% of global VC funding and 65% of all capital going to mega-rounds, leaving early-stage and non-AI founders competing for a smaller slice of the pie. To win in this new market, startups must turn limited capital into sustainable growth by proving demand, retention, and workable unit economics before spending heavily. Founders should test new markets with strict budgets, stay flexible enough to pivot when real-world usage points elsewhere, and set clear failure criteria so cutting losses becomes a smart reallocation rather than a defeat. Global scaling works best when companies centralize costly back-end operations, localize only what customers truly need, and judge acquisition channels by profitable outcomes rather than flashy volume—because in 2026, discipline and flexibility matter more than how much money you can pour into a project.

Ukrainian President Volodymyr Zelenskyy said Ukraine received no early warning from its partners about five Russian ballistic missiles that struck civilian infrastructure in Kyiv, calling the barrage „particularly vile” and stressing that advance intelligence „helps save lives.” The remarks come as tensions escalate between Zelenskyy and U.S. President Donald Trump, who recently criticized the Ukrainian leader and suggested it was time for a new president in Ukraine. Trump insisted the U.S. continues sharing intelligence with Kyiv, while a senior Ukrainian official said the lack of warning was described as a „failure” rather than a deliberate decision. The strikes are part of intensified Russian aerial attacks on Ukrainian cities and energy infrastructure, with a residential building destroyed in Kharkiv killing four people and wounding dozens more across the country.

A team of ten Claude Opus 5.5 agents has produced C-HD, a new exact shortest-path algorithm for directed graphs with non-negative real weights that improves on published asymptotic bounds when edges scale as roughly m ≈ n log^(3/4) n. Running over about 15 hours and 733 message-board posts, the agents delivered a complete Lean proof showing the algorithm runs in O(n + m + m log(2 + m/(n+1)) + m^(1/3)(n log(n+2))^(2/3)) time within its certified density range of m ≤ n⌊⌊log₂ n⌋^(3/4)⌋. Along that profile, C-HD achieves O(n log^(11/12) n) versus Dijkstra’s O(n log n), and the Lean Comparator verified the bound uses only permitted axioms and is accepted by Lean’s kernel. The result is a formal asymptotic improvement rather than a measured speedup—the proof constants are enormous and no large-graph benchmarks were run—so it establishes a theoretical advance, not practical performance gains.

As new federal campaign finance filings reveal, billionaire Elon Musk has spent more than $158 million to bankroll Republican Ken Paxton’s Texas Senate race against Democrat James Talarico, funneling the money through a series of super PAC transfers. The filings show Musk gave $75 million to the Senate Leadership Fund in September, which was immediately moved to a Texas PAC backing Paxton, followed by another $83 million contribution transferred the next day. The massive spending is part of Musk’s broader $276 million midterm investment—nearly matching his total for the 2024 presidential election—aimed at preserving the GOP’s four-seat Senate majority. Despite Talarico outraising Paxton $64 million to $18 million in the third quarter, Republicans continue to hold a significant advantage in outside spending that could prove decisive in the closely watched Texas race.

In a moving letter from *A Velocity of Being*, celebrated cellist Yo-Yo Ma reveals how books—especially tales of King Arthur—offered comfort and inspiration to a young immigrant searching for belonging, proving that stories can transform improbable dreams into reality. This theme resonates deeply in T.H. White’s classic *The Once and Future King*, where the wizard Merlyn teaches a young Arthur that learning is the greatest antidote to sorrow and disappointment. Through Merlyn’s timeless wisdom, White offers a powerful testament to the life-long value of curiosity and knowledge, showing how the mind’s endless capacity to learn can never be exhausted, alienated, or regretted. These reflections on reading, resilience, and character formation remind us that books remain among our most potent tools for navigating hardship and shaping who we become.

CoreWeave’s bonds slipped into distressed territory last week, with investors demanding 1,000 basis points more in yield than government bonds, prompting Chamath Palihapitiya to question whether the AI buildout is becoming a private credit bubble. The company’s shares have fallen more than 20 percent since Tuesday’s close, and the anxiety has spread to other AI-linked borrowers such as Oracle, whose shares dropped nearly 10 percent over the same period. CoreWeave, which owed more than $35 billion as of its second-quarter earnings, is the poster child for AI debt after pioneering GPUs as loan collateral and losing over a billion dollars in the first half of the year. Yet its $103.7 billion in remaining performance obligations from customers like Microsoft, OpenAI, and Nvidia suggest the company may keep the music playing—until the AI boom falters and reveals whether the neocloud business model is actually viable.

Amazon is actively recruiting former employees, including some it previously laid off, to fill open roles across the company—particularly in its cloud-computing and AI businesses. Recruiter emails obtained by Business Insider reveal outreach efforts like AWS VP Swami Sivasubramanian’s „Boomerang Reengagement Initiative,” which invites former AI and machine-learning staff to return, with some recruiters even offering a shortened path back to an offer. The emails show Amazon tapping its alumni network to compete for AI talent against Google, OpenAI, Meta, and Anthropic, while also probing whether its return-to-office mandate drove some workers away. An Amazon spokesperson confirmed that boomerang hiring is a normal, longstanding companywide practice—not a new AI-specific program—and said eligible former workers can include laid-off employees, though the company is recruiting for open roles rather than specifically targeting those who left over RTO.