Vals, backed by Andreessen Horowitz, is looking to become the gold standard for AI benchmarking

techcrunch.com Sep 20, 2026

Vals, backed by Andreessen Horowitz, is looking to become the gold standard for AI benchmarking

Curated by Shubham · TechCapsules · Sep 20, 2026

AI-summarised brief · reviewed before publication

Vals, a San Francisco AI‑benchmarking startup founded in 2024, raised $40 million in a Series A led by Andreessen Horowitz after an earlier seed round from 8VC and Bloomberg Beta. Co‑founder Rayan Krishnan, a former Palantir intern and Stanford AI lab student, says the company was created to address outdated public benchmarks that can be gamed by model developers. Vals offers private, industry‑specific tests—covering law, finance, coding and emerging areas such as mental‑health, cybersecurity and the law of armed conflict—rather than generic knowledge exams. Clients pay for these evaluations to identify performance gaps and mitigate negative outcomes. Revenue has reportedly grown eightfold in a year, and staff expanded from eight to 25, with plans to add more and serve federal agencies in the near term.

💡 Why It Matters

· By privatizing test data and targeting real‑world tasks, Vals forces AI firms to substantiate functional value instead of merely chasing headline scores.

AI Capsules Funding Startups

Read full article on techcrunch.com

Figure AI Helix 2.5 Enters 30 Homes Cold: Index Pretraining Yields Sixfold…

On September 17, 2026, Figure AI announced that its Helix 2.5 neural network successfully completed 56% of 420 household tasks in 30 Bay Area homes it had never seen before, using a single pretrained checkpoint from the Index dataset. The same architecture trained from random weights achieved only 9%. The experiment involved three tasks—living‑room tidying, towel folding, and bed making—without any data collection or fine‑tuning on the new environments. The results demonstrate a sixfold improvement attributable solely to Index pretraining, not hardware or model size.

💡 The finding shows that large‑scale human‑behavior data can unlock reliable zero‑shot performance in real homes, potentially accelerating the deployment of autonomous domestic robots.

taipeitimes.com · Sep 20

AI could reverse political polarization with probability-based centrism

The article argues that large language models, while often viewed as divisive, may unintentionally foster political consensus by shaping how people learn. It traces polarization’s roots to media fragmentation, noting how the 1970s cable boom and the internet’s rise created echo chambers that eroded shared narratives. The piece contrasts this with earlier broadcast eras, where a handful of neutral anchors like Walter Cronkite provided a common informational baseline. By highlighting AI data centers as a potential bridge, the author suggests that algorithmic content curation could revive a national dialogue.

💡 If AI can recalibrate public discourse, it offers a rare tool to counteract deepening partisan divides that have long stifled bipartisan cooperation.

moneycontrol.com · Sep 20

Claude is building its successor as it takes on 26 percent of…

Anthropic announced that its flagship model, Claude, now leads approximately 26 percent of the company’s research and development activities, a sharp rise from zero earlier this year. The figure reflects tasks where Claude can execute most of the work from high‑level instructions while human researchers supervise and set objectives. Anthropic maintains that Claude remains non‑autonomous, with humans deciding research goals and evaluating outcomes. The company disclosed the metric to provide transparency on AI involvement in its own development and to address concerns about recursive self‑improvement.

💡 By quantifying Claude’s growing role, Anthropic offers a concrete benchmark for industry observers to gauge how quickly AI systems are taking over their own creation, informing safety and policy discussions.

recombee.com · Sep 20

The AI (R)Evolution in the Media Industry

The media sector is undergoing a rapid AI‑driven transformation, with personalization emerging as its core catalyst. A recent PwC report projects AI will dominate content creation, distribution and user‑experience design by 2030, reshaping how audiences engage with media. Professionals cite AI’s impact across production efficiency, cost reduction, advanced CGI, and virtual production, while AI‑enhanced cameras now automatically track athletes during live sports, delivering dynamic close‑ups. Fraud‑detection algorithms combat ad fraud and copyright violations in real time, safeguarding creators and consumers. Meanwhile, churn‑prevention tools analyze behavioral data to flag at‑risk subscribers, enabling proactive retention strategies. Despite these gains, experts warn that AI’s power carries ethical risks, prompting calls for regulatory frameworks to ensure responsible deployment in the industry.

💡 AI is turning media consumption into a hyper‑personalized experience that can lock in audiences and protect revenue streams, while the looming ethical concerns demand immediate policy attention.

mediaite.com · Sep 20

Trump Asks Americans to Vote on Name Replacement for AI in Puzzling…

President Donald Trump used his Truth Social platform on Saturday to launch a public poll asking Americans to choose a new term to replace “artificial intelligence.” He criticized the existing label as “inaccurate” and “very ineloquent,” proposing three alternatives: Super Intelligence, Extreme Intelligence, and Supreme Intelligence. The post referenced the 1955 origin of the term by Stanford professor John McCarthy and noted Trump’s frequent use of AI for his own social‑media content. The poll appears amid heightened industry debate, highlighted by Anthropic CEO Dario Amodei’s call for slower AI development and warnings from researchers about existential risks. Trump dismissed such concerns as a “hoax,” reiterating his stance that the United States must outpace rivals like China in the emerging technology race.

💡 By turning a technical naming dispute into a political poll, Trump injects partisan framing into AI discourse, potentially swaying public opinion and influencing future regulatory approaches.

moneycontrol.com · Sep 20

How OpenAI’s copyright battle with The New York Times reached a new…

In a new filing, The New York Times alleges that OpenAI copied millions of copyrighted articles without permission while training its AI models, accusing the firm and its partner Microsoft of “transgressive and deceptive” methods to bypass paywalls. The lawsuit, part of a broader wave of copyright actions against AI developers, seeks billions in damages and argues that OpenAI’s data collection was driven by commercial motives, citing a 2017 statement by co‑founder Greg Brockman about financial rewards. The filing also references internal ChatGPT leadership warnings that AI poses an “existential threat” to publishers. OpenAI maintains its use of publicly available material qualifies as fair use, likening it to search‑engine indexing and asserting its products do not substitute for news content. The case remains unresolved, with Microsoft distancing itself from an employee’s comment labeling the practice as theft.

💡 The dispute could set a precedent for how AI systems may legally source copyrighted content, potentially reshaping the balance between technological innovation and media owners’ control over their work.