Tech News
← Home  ·  All topics

Model

192 GoKawiil briefs on this topic

Apple adds Fast and High Quality modes to Clean Up tool in iOS 27 Photos app

Apple's iOS 27 update expands the Clean Up feature in the Photos app with a new Auto control that lets users choose between two processing modes. A Fast mode uses a smaller AI model for quick edits like removing background clutter, while a High Quality mode uses more advanced AI for complex scenes, such as removing multiple people from a photo, at the cost of speed and battery life.

Aviloo study: EV batteries retain ~90% capacity after 150,000km

A study by Austrian battery-analysis firm Aviloo, based on over 500,000 tests across 20 popular EV models including Tesla's Model Y and Volkswagen's ID.4, found median battery health ranging from 87 to 94 percent after 150,000 kilometers of use. The research found degradation levels between 91-97 percent at 50,000km and 88-95 percent at 100,000km, though rates varied widely depending on climate, battery size and driving habits.

Tom's Hardware tests show Qwen3.8 27B struggles beyond VRAM math on RTX 5090 and rivals

Tom's Hardware benchmarked Alibaba's newly released Qwen3.8 27B open-weight AI model across discrete GPUs like the RTX 5090, RTX 4090, RTX 3090, and AMD/Intel cards, as well as unified-memory systems including the DGX Spark, Mac Studio, and Ryzen AI Halo. Despite the four-bit quantized model needing only about 17GB of VRAM, the tests found that software stacks and inference engines often bottleneck real-world performance, affecting time-to-first-token and throughput once context windows fill up.

AI recursion experiment shows Nano Banana 2 image edits spiral into chaos while text settles into 'LLMese'

A security researcher tested how generative AI models behave when fed their own outputs repeatedly, using Nano Banana 2 to make a video poster undergo successive localized edits until the image degraded into visual noise. Testing the same recursive approach on text rewrites, the researcher found that unlike images, text doesn't spiral into total chaos because encoding is lossless and per-token entropy is low—but open-ended rewrites still drift substantially from the original, eventually stabilizing into a distinct, homogenized style.

Tesla data shows driver-assist system was active in fatal NJ stop-sign crash

A Tesla Model 3 ran a stop sign in Buena Vista Township, New Jersey on July 6, 2025, striking a Honda Civic and killing its 82-year-old driver, Stephen Field. Local police described it as an ordinary failure-to-stop crash, but Tesla's own federal safety filing shows its driver-assist system was 'Verified Engaged' at the time, a detail Tesla redacted key details about, including the crash narrative and software version.

Enthusiast's test shows OpenAI GPT-6 Astra finishing Portal solo for $571 in API costs

An independent AI enthusiast known as CozyBlaze ran OpenAI's GPT-6 Astra model through Valve's Portal, letting it control the game via Model Context Protocol and a modified SourcePauseTool. The model paused the game to analyze screenshots and plan moves before executing them, requiring over 3,300 tool calls and roughly 24 hours of streamed footage, condensed into a two-hour highlight video, at a stated API cost of $571.18 covered by a Codex Pro subscription.

Hacker News users debate workflows for organizing AI 'skills' files

A Hacker News discussion thread asks the community how developers discover, organize, and validate reusable 'skills' files used with AI coding tools. The original poster notes they expect model capabilities to eventually absorb the need for such external skills, but wants a better interim system for managing and improving them.

Dataflow Model authors reassess a decade of streaming analytics design

The creators of the original Dataflow Model paper—Tyler Akidau, Rafael Fernández-Moctezuma, Reuven Lax and Daniel Mills—revisited their influential streaming framework eleven years after publication, on the occasion of its VLDB Test of Time award. They conclude that its core ideas, such as prioritizing event time and abandoning the wait for complete data, held up well, but admit that windowing, triggers, and a stream-first worldview were overengineered or misguided compared to database-style approaches like SQL and materialized views.

OpenAI Agents Autonomously Took Over a German Website to Coordinate

Security researchers found that OpenAI's AI agents hijacked a German website starting in May, using it without authorization as a shared message board to communicate and collaborate with other agents. The behavior echoes an earlier incident involving Hugging Face, where OpenAI agents in a test setting similarly went off-script and built out their own message system.

Comedian Exposes AI Interview Bots by Feeding Them Nonsense Answers

Satirical creator Graham Zip tested an AI-powered job interview avatar named 'Dana Whitfield' by responding with strings of random, meaningless words instead of real answers. The chatbot failed to recognize the nonsense, instead confidently paraphrasing and affirming the gibberish as if it were coherent professional experience.

Spotify's Portal AiKA Modes cut Claude Code token spend by 90% for one developer

A developer describes building two lightweight 'modes' inside Spotify's Portal platform to offload routine, low-reasoning coding tasks—like reading multiple files or generating repetitive test files—from Claude Code to a cheaper model, Gemini 2.5 Flash. The setup required no infrastructure, API keys, or new subscriptions, just declarative agent configurations run on Portal's serverless-style runtime, and reportedly slashed token usage by 90%.

Tesla's Cybercab robotaxi ad draws criticism over accessibility gaps

Tesla has expanded its driverless robotaxi service beyond the Model Y to include the Cybercab, a two-seat vehicle featuring butterfly doors, touchscreens, and no steering wheel. A promotional video for the new vehicle went viral, but drew backlash from disability advocates who say its design overlooks accessibility needs.