gpt

An AI Zeitgeist study surveyed seven GPT releases with 18 questions and 15,120 answers, finding they disagreed on humanity's biggest threat—climate change for GPT-5 through GPT-5.6 Sol, AI misalignment for GPT-5.4/5.5, and nuclear war for GPT-6 Astra—while agreeing AI agents will most change daily life.

A claim surfaced that the upcoming next-generation GPT model used 10,000 agents working for about four days to prove the first Millennium Prize problem, with the model expected to be released soon.

FARO is a function-space adaptive optimizer that assigns each parameter block a geometry based on its counterfactual return and reallocates the per-block update budget via cross-block equal risk contribution, matching hand-designed recipes like NorMuonAndAdam while providing robust small loss gains on GPT-style training.

The author shares their ordeal of buying credits for AI services such as GPT and Gemini: after repeatedly running into risk-control blocks with Japanese-region Apple gift cards and a China Merchants Bank Visa card, they finally topped up successfully using a Gate card (buying USDT via C2C), and outlines the shortest payment route.

A speculative report suggests a new OpenAI model, possibly GPT-6.2, may be on the horizon, though the claims are unconfirmed and should be treated cautiously.

A user offers an objective take on the Kimi 3.1 Pro controversy, arguing that despite the heavy criticism over Kimi's pricing and models, the actual user experience is still solid — and that its content design and frontend capabilities even outperform current GPT, which is better suited for debugging and verification.

Thariq (@trq212) shares UI explorations for his personal elemental fighting game prototype, using Claude and GPT for image generation while noting frustrations with agent-generated UI adding too many details.

Artificial Analysis intelligence index v5 is drawing criticism for introducing private tests on private datasets that appear to favor closed models like GPT and Google's while significantly downranking open-weight Chinese models such as Kimi 3.1 Pro and DeepSeek 4.1 Flash, coinciding with Google's Carbon 4 topping the index.

A warning post advising users not to swap models in Codex's Dreaming V3 memory system, since prompts are carefully calibrated for the default GPT-5.6 Luna and Terra models.

GPT-6.1 Terra 发布，OpenAI 称其在快速、能干的计算机使用（computer use）方面实现新突破，能理解屏幕内容、跨应用执行操作，并以更低价格提供更高效的工作能力。

Wiz disclosed PHP2Shell, two critical decade-old PHP vulnerabilities that researchers turned into working exploits in about a week with help from Gemini and GPT models, prompting emergency patching and raising questions about AI-driven exploitation of legacy code.

A tweet claims DeepSeek v4.1 Flash 1017 performs comparably to GPT 5.6 Sol, describing it as an at-home equivalent.

Yohei Nakajima highlights research by Jasper Dekoninck showing that GPT-6.1 Sol can detect errors in arXiv papers' main theorems, finding ~6% of pre-August 2026 papers contain false main theorems versus only 1.5% in recent months. The post advocates for broader use of AI tools like exscienceAI in academic peer review.

一条 Thread Reader 聚合帖宣称 GPT 6.1 Sol 默认上下文窗口达到 500k，但未提供官方来源佐证，可能属于传闻或推测性消息。

Google's Gemini models have largely disappeared from the top of the LM Arena coding leaderboard, with Claude and GPT models now dominating, prompting questions about whether Google is shifting priorities toward multimodal AI, search, and agents rather than competing for top coding-model status.

该论文研究了GPT模型在帮助润色学术论文时出现的"防御性写作"现象——即无依据地收窄或撤回作者论点，发现这主要源于模型在为预期的审稿人写作，而非修正过度声明，且AI审稿人倾向于给防御性改写更高评分，可能形成正反馈循环。

The author tasked Claude (Fable 5.1) and GPT (GPT6 Astra) with independently building a production-grade C++ 3D game engine, first scoping the task and writing requirements documents to let the models work autonomously. The article explores whether AI can carry out game engine development end to end, and to what extent AI can replace an experienced engineering team.

作者提出一个衡量AGI的新标准：一个通用大模型能够仅凭纯视觉/音频输入和人类相同的控制方式，实时、可靠地在多种专业电竞比赛中与职业选手同场竞技，且具备真实反应、机制理解、策略竞争性和极低幻觉错误率的能力。

An OpenAI researcher announces their departure after seven years, reflecting on the evolution from GPT-2 to GPT-4, ChatGPT, CLIP, DALL·E, and Codex, and sharing plans to focus on family.

Artificial Analysis' new Coding Agent Index shows Claude Sonnet 5.5 in Claude Code leading at 68, but Gemini 4 Argon in Antigravity CLI (64) is close behind at less than half the cost, while GPT-6.1 Sol in Codex scores 63 at only $1.04 per task.