This Week in Tech: Regulators Take On Google, a New Coding Champion Emerges, and Chipmakers Keep Breaking Records
- Get link
- X
- Other Apps
This Week in Tech: Regulators Take On Google, a New Coding Champion Emerges, and Chipmakers Keep Breaking Records
⚖️ EU Forces Google to Open Android to Rival AI Assistants
The European Commission this week adopted binding requirements under the Digital Markets Act that will force Google to open Android to competing AI assistants and share portions of its search index with rival developers, including AI companies building competing products. The order marks one of the most consequential regulatory actions taken against the company since the DMA came into force, and it directly targets the default-assistant privileges that have long kept Google Assistant and Gemini positioned ahead of alternatives on Android devices.
Under the ruling, search data sharing is set to begin in January 2027, with full Android interoperability requirements due by July 2027, giving Google roughly a year to restructure how third-party assistants integrate with its mobile operating system. Brussels regulators framed the decision as necessary to prevent Google from using its platform ownership to entrench its own AI products at the expense of competitors trying to reach the same 3 billion active Android devices worldwide.
The timing is notable: it lands just as Google's own Gemini 3.5 Pro has reportedly slipped in internal testing for a third time, falling short on coding and complex reasoning benchmarks. That stumble, combined with the new interoperability mandate, could open meaningful room for rival assistants to gain ground on Android over the next 18 months.
💻 Moonshot AI's Kimi K3 Tops the Coding Leaderboards
Chinese AI lab Moonshot AI released Kimi K3 this week, and the model immediately climbed to number one on Arena.ai's Frontend Code Arena leaderboard with a 76 percent pairwise win rate against competing systems. The model also scored 88.3 on Terminal-Bench 2.1, a benchmark designed to measure how well an AI system can operate autonomously inside a command-line environment, completing multi-step engineering tasks without human intervention.
The result is significant because it underscores how quickly the gap between Western and Chinese frontier labs has narrowed on specialized tasks like software engineering and autonomous agent workflows. Kimi K3 is being positioned by Moonshot AI less as a general chatbot and more as a dedicated coding and agent specialist, a strategy that appears to be paying off as enterprise developers look for models that can reliably execute multi-step technical work rather than simply answer questions.
Analysts tracking the leaderboard note that Kimi K3's showing arrives at an awkward moment for some Western labs still finalizing their next flagship coding releases, adding pressure on the broader industry to demonstrate that upcoming models can match or beat the new benchmark.
🏭 TSMC Posts Blockbuster Quarter, Doubles Down on Arizona
Taiwan Semiconductor Manufacturing Company reported second-quarter net income up 77.4 percent year-on-year this week, comfortably beating analyst expectations and prompting the company to raise its full-year revenue guidance to more than 40 percent growth. High-performance computing chips, the category that includes AI accelerators for Nvidia and other customers, now account for 66 percent of TSMC's quarterly revenue, a sign of just how central AI demand has become to the world's most important chipmaker.
Alongside the earnings beat, TSMC confirmed a further $100 billion in investment for its Arizona manufacturing campus, extending a build-out that began earlier in the decade and that U.S. officials have repeatedly cited as a cornerstone of domestic semiconductor policy. The additional funding will expand advanced packaging and logic fabrication capacity specifically earmarked for AI chip customers.
The results land amid a broader debate over whether AI infrastructure spending across the industry, from data centers to fiber networks to power generation, is being matched by proportional returns. TSMC's numbers offer at least one clear data point suggesting that demand for the physical hardware underpinning AI remains, for now, undiminished.
🎙️ OpenAI Widens Rollout of GPT-5.6 and Full-Duplex Voice
OpenAI is moving its GPT-5.6 family, internally named Sol, Luna, and Terra, beyond limited preview and into broader public availability, following additional testing requested by U.S. government reviewers. The staged rollout suggests the company is being more deliberate about deployment timelines for its newest frontier models than it has been with past releases, likely reflecting heightened regulatory attention on model safety evaluations before wide release.
Alongside the model rollout, OpenAI introduced GPT-Live-1 and a smaller GPT-Live-1 mini variant for ChatGPT's voice mode, giving both paid and free-tier users access to full-duplex conversational audio. Full-duplex capability allows the system to listen and respond simultaneously rather than waiting for a speaker to finish, closing much of the latency gap that has made AI voice assistants feel noticeably less natural than human conversation.
Taken together, the two releases point to OpenAI prioritizing conversational fluidity and broader access this quarter, even as competitors like Moonshot AI make gains in specialized coding benchmarks. The company has not detailed pricing tiers for GPT-Live-1 mini, though early reports suggest it will be included in the free ChatGPT tier with usage limits.
Sources
- Get link
- X
- Other Apps
.webp)
Comments
Post a Comment