Benchmark

Australia's central bank has raised its benchmark interest rate to its highest level in around 15 years, according to the Financial Times. The move means more expensive borrowing for Australian households and businesses, and is aimed at curbing inflation. Reactions so far focus on the pressure this adds to mortgages and the broader cost-of-living squeeze.

Google Research has open-sourced RRSI, a framework allowing AI agents to refine their own evaluation harness while guarding against overfitting. Announced via MarkTechPost, the release lets developers inspect and build on the underlying code. The announcement is drawing attention from AI practitioners interested in agent self-improvement methods that remain reliable rather than gaming their own benchmarks.

A self-play reinforcement learning bot has beaten a strong human player at StarCraft: Brood War, one of the most demanding competitive strategy games in existence. The result, shared in a video, is drawing attention because Brood War's complexity has long made it a harder benchmark for AI than games like Go or chess, following in the footsteps of earlier research efforts such as DeepMind's AlphaStar project.

Japan's Nikkei index closed 0.8% lower, with losses led by automakers and steel companies. Market watchers pointed to sector weakness in autos and steel as the main drag on the benchmark. Traders were watching the session closely for what it signals about Japan's export-heavy industries.

Yahoo Finance is asking whether earning a six-figure salary still counts as the benchmark for financial success. With inflation, higher living costs and rising salaries, $100,000 no longer buys what it once did, prompting debate over whether that threshold remains meaningful. Commenters are weighing in on what income level today genuinely signals financial security and success.

Benchmark has closed Eleven Mortgage, its wholesale lending division, and says it will concentrate on its retail mortgage business going forward. The move means wholesale borrowers and brokers who worked with Eleven Mortgage will need to find other lenders. The company has not detailed how many employees are affected or the reasons behind the decision, which comes as mortgage lenders continue reshaping their operations amid a difficult housing market.

Commentary argues that the competition facing Apple's anticipated two-screen iPhone lineup will not come from Samsung, Google or other Android foldable makers, but from elsewhere entirely — the implied point being that existing Android foldables have not defined the category the way Apple could. The piece fuels ongoing speculation about how Apple will approach its first foldable and what benchmark it must beat.

Solar software company Aurora Solar has launched Aurora Solar Trust Signal, a quarterly tracker measuring homeowner trust in the solar industry. The company says it is the first tool of its kind in the sector, aiming to give installers and manufacturers regular data on how consumers perceive solar energy and what drives their confidence or skepticism when adopting it.

A new evaluation from Fig reports that frontier AI models including Astra and Opus 5.5 perform unevenly — 'jagged' — across agentic tasks, excelling at some web browsing challenges while struggling on others, including robotics workloads. The findings suggest that no single model yet dominates all agentic domains, and that benchmark scores on one task type may not predict performance on another.

Mortgage and refinance rates were updated for Monday, September 28, 2026, with refinance rates remaining lower than purchase rates. The daily rate tracker gives borrowers a benchmark for what lenders are offering, and the continued gap between refinance and purchase pricing is the notable takeaway for homeowners weighing a refinance against current purchase terms.

Costco's stock is being framed as the new benchmark for recession-resistant retail, with commentary arguing the warehouse club's membership model and bulk-value pricing position it to hold up better than traditional retailers if the economy slows. Analysts and investors are weighing whether that resilience justifies the premium the shares trade at, as shoppers keep seeking value amid persistent cost-of-living pressure.

Microsoft Research has introduced Quine, a multimodal world model designed for biology, unveiled through a report by Unite.AI. The system is described as a world model, meaning it aims to simulate and understand biological processes rather than just a single task. Details on its capabilities, benchmarks, and availability remain limited in initial coverage, and Microsoft has not yet been widely quoted on plans for the model.

Early over/under win total projections for the 2026/27 NBA season are out for the Northwest Division, covering the Nuggets, Thunder, Timberwolves, Trail Blazers and Jazz. The lines give fans and analysts a first benchmark for next season, prompting debate over which teams will exceed expectations and which will fall short.

vLLM has added text watermarking based on the Gumbel-max trick: pseudo-random noise derived from a secret key and the last four tokens is applied without altering the generated text. The developer reports quality gaps of under 2 points on GSM8K, MBPP and IFEval benchmarks, with a signal that can be detected without access to the underlying model. The feature is drawing attention among people following AI-generated content detection.

Travel industry publication Skift reports that Europe has become the key testing ground for artificial intelligence in travel, because the continent's many languages, national rail and airline systems, and differing regulations make cross-border journeys the hardest case for AI trip planning and booking tools. If AI assistants can handle a multi-country European trip, the argument goes, they can work anywhere.

Retail analysts say shoppers increasingly use artificial intelligence tools to find the best deals, and that this behaviour is becoming a benchmark for customer experience. Retailers that make offers easy for AI assistants to discover and compare are seen as better positioned, while those with fragmented pricing or poor product data risk losing bargain-hunting customers to more AI-friendly competitors.

Vietnam's Ministry of Science and Technology has outlined requirements for importing used technology production lines: the equipment must operate at a minimum of 85% efficiency with no more than 15% consumption relative to comparable new technology. The standards are intended to block obsolete machinery and ensure imported second-hand industrial lines meet modern energy and performance benchmarks.

Wall Street's main indexes were holding steady in morning trading, with little movement across major benchmarks as investors weighed the market outlook. The quiet session left stocks near recent levels, and market watchers are watching for signals on where US equities head next.

Mercedes have overtaken Ferrari to become the most successful manufacturer in Formula 1 history, according to RacingNews365. The German marque has now surpassed the Italian team on the all-time constructors' records, ending Ferrari's long-standing position at the top of the sport's historical standings. It marks a major symbolic shift in Formula 1, where Ferrari had been the benchmark team for decades.