An analysis argues that the cost of running machine learning models is dropping by orders of magnitude annually, driven by GPU efficiency gains that double roughly every two years—a pace not seen since early Moore's Law. The piece distinguishes proprietary models like GPT-6 Astra from open-weight models such as GLM-5.3-flash, noting that hosted and locally-run versions improve at different rates, with per-token pricing for frontier models not falling as consistently as costs for smaller models.
jyn.dev
· 2026-09-23
Scientists developed a surface treatment method aimed at improving both efficiency and durability of large-area perovskite solar modules, according to a report published via Nature. The approach targets scaling up perovskite photovoltaic technology while addressing the stability issues that have limited its commercial viability.
nature.com
· 2026-09-23
Unreal Labs introduced Unreal Agent, a harness for running AI agents that it says delivers up to 40% cost savings compared to OpenAI's Codex and up to 20% savings compared to Pi, without hurting performance on production workloads and coding/science benchmarks. The company says the harness manages tool calls asynchronously, removing the need for the underlying model to handle waits, polls, and heartbeats itself.
unreallabs.ai
· 2026-09-22
MIT Technology Review's latest list of under-35 innovators highlights researchers and founders working on climate tech, including Jae-Won Chung's software for measuring AI models' energy consumption, Jing Wei's AI-based pollution tracking, and Zhonghua Zheng's city-specific climate models. The list also features Mohammad Alkhadra of Lithios and Benjamin Mowbray of Rock Zero, two startups developing faster, cheaper methods to extract lithium from brine and hardrock ore respectively.
technologyreview.com
· 2026-09-17
Processes originally created to solve a specific business problem can outlive their usefulness, becoming bottlenecks rather than safeguards. Leaders frequently continue enforcing these procedures long after the original risk has faded, mistaking the fact that a process once worked for proof it remains necessary today.
entrepreneur.com
· 2026-09-15
Volkswagen unveiled and actually drove a new prototype EV called Mission Efficiency, taking it nearly 800 miles from Wolfsburg to Vienna with just one charging stop. The car averaged about 279 MPGe on the trip, using the motor and battery from the ID. Polo housed in an aerodynamic, teardrop-shaped body with a drag coefficient of just 0.158.
techcrunch.com
· 2026-09-14
Volkswagen unveiled the Mission Efficiency, a near-production electric coupe prototype built on its MEB+ platform that shares components with the upcoming ID. Polo and ID. Cross models. The teardrop-shaped car achieved a drag coefficient of 0.158, which VW says makes it the world's most aerodynamic road-legal vehicle, and completed a 1,278 km trip from Wolfsburg to Vienna on one charge of its 54.9 kWh battery, arriving with 164 km of range left.
theverge.com
· 2026-09-14
Chipotle developed a High-Efficiency Equipment Package after frontline workers flagged that slow prep and grilling were hurting throughput and consistency. The package, built through rapid R&D iteration and in-restaurant testing, includes a three-pan rice cooker, dual-sided plancha griddle, high-capacity chip fryer, and produce slicer, cutting prep time by hours per restaurant daily. The innovation helped land Chipotle the No. 2 spot on a Best Workplaces for Innovators 2026 list.
fastcompany.com
· 2026-09-08
Startup Newlight tested a retrofit hydrogen-injection system on a 650-foot Lomar Shipping bulk carrier during an 8,500-nautical-mile voyage from Singapore to Ghana. The system, which precisely times hydrogen injections alongside diesel combustion, reduced fuel consumption by 24% and cut carbon dioxide emissions by 28% during the month-long trip. Newlight says the retrofit can be installed in under two weeks without dry-docking a vessel.
techcrunch.com
· 2026-09-01
A researcher built and trained a small transformer model from scratch in 1.5 hours on a single RTX 5090 GPU, achieving 44% accuracy on the ARC-AGI-1 public evaluation set and 7% on ARC-AGI-2. The model, an upgraded and cheaper version of a previous open-source release, matches the performance of existing test-time-training approaches like TRM and HRM while using far less compute.
mvakde.github.io
· 2026-09-01
A commentary argues that companies are being pushed to spend heavily on artificial intelligence not because it guarantees financial returns, but because rivals adopting the technology will force everyone else to follow suit just to remain competitive. The piece frames AI spending as a defensive necessity rather than a growth strategy.
wsj.com
· 2026-08-31
After a March 2025 executive order targeted the Community Development Financial Institutions Fund for elimination, B:Side Capital CEO faced an existential threat to the SBA-backed lender's operations. Rather than scrambling for a new strategy, the CEO relied on a previously established written code of decision-making principles to guide the organization through the weekend crisis and present a plan to staff and board by Monday.
entrepreneur.com
· 2026-08-27
GoKawiil