AI — Page 2

Wiz found missing Supabase controls let outsiders read and write production data before Meta bought the AI-agent forum.

This one is close on aggregate, but the split decision tells you exactly where each model lives. Seedance wins when the brief demands harder camera choreography and stricter prompt obedience; Luma still has the edge when atmosphere and intimate mood need to carry the shot.

One model consistently understood the assignment; the other mostly produced attractive approximations. Across layout, depth, and typography, this matchup wasn’t especially close.

This matchup turns on instruction discipline versus presentation polish. Cohere-command-a steals one business-writing round, but grok-4.3 is the more reliable model where exactness actually matters.

Z.ai is using open weights and lower token costs to pressure Anthropic and OpenAI where enterprises feel the bill.

This one is close on points, but not in character. Bytedance Seedance V1.5 Pro is the stricter prompt-follower on human detail, while Happy Horse 1.1 wins by delivering the more convincing, better-shot video when mood and camera language matter most.

Brett Hurt's Love Conquers Fear is not another AI book. It is a founder memo for the species, and its argument is simple: abundance is technically possible, but fear can still ruin everything.

Tom Brown has taken the lead in talks with Commerce after a June 12 order forced Anthropic to shut off its top models globally.

The lab claims its MoA setup beats Opus 4.8 and GPT 5.5 on an unreleased HermesBench, but the benchmark is not public yet.

OpenAI gave the evaluator raw chain-of-thought and a railfree model, but METR says the results were too unstable for a robust time-horizon read.

This one isn’t close: Fibo wins by being the model that actually obeys the brief when the brief gets specific. Bagel can make pretty images, but across all three tests Fibo delivered the stronger read on scene logic, action, and physical detail.

The new Sol, Terra and Luna lineup is headed first to trusted partners in Codex and the API before a wider release OpenAI says is weeks away.

Naveen Rao's hardware startup is using a software model to test whether physical dynamics can carry real AI workloads before its chips exist.

Azeem Azhar's team says its bottom-up model strips out double-counting and measures end-customer AI spending across the stack.

CS2-10k promises 10,000-plus hours of first-person video with synchronized controls, but the full Hugging Face upload is not live yet.

This matchup wasn’t close once motion direction, subject fidelity, and scene logic were put under pressure. Bytedance Seedance V1.5 Pro Image To Video can produce attractive frames, but Happy Horse was the one that actually followed the briefs.

The open release is a software simulation, not a chip, but it gives Rao's $475 million energy-efficiency bet its first public benchmark.

Sam Altman told staff the federal government asked OpenAI to limit GPT 5.6 access before wider release, The Information reported.

This wasn’t close. Across three very different prompts, Cosmos 3 Super was the model that more reliably understood the brief, while AuraFlow kept producing images that looked lively but missed crucial instructions.

The endpoint prices audio generation at $0.075 per minute and gives developers one API surface for text, voice, references, and image-guided sound.

One model was consistently cleaner, sharper, and better at following instructions; the other kept leaving small but consequential mistakes on the table. Across coding, transformation, summarization, and business writing, this matchup wasn’t especially close.

YC's Spring 2025 company pairs a Harvard Law CEO with an MIT engineer and says legal intake can start inside chat apps.

The June 24 release targets ChatGPT's default workhorse model, with OpenAI pushing it deeper into recommendation and decision queries.

The Claude maker says the April-to-June campaign produced 28.8 million exchanges and targeted coding and agentic reasoning.