X

前段时间参与了 Doubao-Seed-2.0-lite 0428 内测。 这个版本升级，增加了音频理解，能同时支持图片、视频、音频、文本四种输入，成为豆包大模型家族首款全模态理解模型。 除了全模态理解，据说 Agent、Coding、GUI 能力这次也都有明显提升。 拿 API 做了一些测试，分享几个场景：前端动效复刻、视频Hooks建议、字幕识别等 案例见后续 Thread

How does Exa compare to Google for training LLMs to search? In this blog post, we find that LLMs using Exa during reinforcement learning reach higher performance with 70% less training compute. exa.ai/blog/rl-search-outcom… Video

Claude Code weekly limits are increasing 50%, now through July 13. Live now for all Pro, Max, Team, and seat-based Enterprise users.

🧪 10M Token Bill Challenge · How to Enter Prize: 30 winners, each receiving 10M OpenRouter credits Timeline: Opens May 7 · Closes May 17 · Winners announced May 20 Submit on X Quote-retweet this thread with your token bill and a one-line takeaway, tag @OpenSquilla, and use #10MTokenChallenge. Tracks: 🥇 Task Reproduction: Pick 3 out of 5 official reference tasks — build a web/mini app/plugin, create a data dashboard, generate an infographic, make a slide deck, or complete a research analysis — and compare OpenSquilla vs another agent 💰 Best Savings Case: Use a real task from your daily workflow and show who saves more while maintaining quality 🐛 Quality Bug Report: Submit a high-quality GitHub issue Comparison rules: ✅ Leverage OpenSquilla with cheap tokens to outperform Opus 4.7 agents ✅ Run each task at least 3 times per side ✅ Record input tokens / output tokens / $ cost / pass or fail ✅ Include your prompt, result table, quality notes, and raw logs We are moving beyond token burning. We incentivize cost-effective, useful, and reproducible agent outcomes.

We released experimental MTP Qwen3.6 Unsloth GGUFs! Qwen3.6 27B MTP now runs at 140 tokens/s. Qwen3.6 35B-A3B MTP gets 220 tokens/s generation on a single GPU. Qwen3.6 27B and 35B-A3B have >1.4x speed-up over the original GGUFs without any change in accuracy. Guide + GGUFs + Benchmarks: unsloth.ai/docs/models/qwen3… In terms of average speedup, we see a 1.4x for dense models at draft tokens = 2 and for the MoE around 1.15 to 1.2x. We do not recommend more than 2 draft tokens because the acceptance rate drops precipitously from 83% to 50% with 4 draft tokens, and the forward passes for MTP become less beneficial. Use `--spec-type mtp --spec-draft-n-max 2` Thanks to Aman for github.com/ggml-org/llama.cp…!

Instead of stitching models together, SenseNova U1 runs everything inside one unified system. Built on NEO-Unify, it treats text and images as part of the same space. No switching between components. No translation gaps. Just one model that understands and creates at the same time.

🚨 FIRST GEMINI 3.2 PRO OUTPUTS From yours truly. I'll share more today, I can test your prompts too if you guys want. I guess 3.2 Pro is also coming with Flash :)

Install ntn, the Notion CLI. It brings the entire Notion API to your terminal, plus everything you need to build and deploy Workers. Built for humans and coding agents alike. Install with: curl -fsSL ntn.dev | bash Video

Introducing: enrich-excel Hermes Agent or @openclaw can now enrich any Excel sheet with live web data. > uses our free Search and Fetch to find missing data > writes results directly into the running Excel via AppleScript All in just a few seconds and one Skill. Video

To add some clarity: you don't pay extra. It's the same subscription, same price per month. What's new our sub now covers two separate pools: · Interactive → sub limits, unchanged · Programmatic → new $20–$200 included(!!) credit, metered at API rates Community note: This is missing key context: programmatic use no longer draws from subscription limits. It uses a separate $20-$200 monthly Agent SDK credit, then extra usage is billed at standard API rates, so heavy sub users may get far less usage than before. support.claude.com/en/articles/15… platform.claude.com/docs/en/about-… support.claude.com/en/articles/12…

Godfather of AI: "If you sleep well tonight, you may not have understood this lecture." This 47-minute lecture is the best thing I saw about AI in the last few months. It will definitely help you understand how it actually works and where it's going. Geoffrey Hinton built the neural networks behind every AI alive, then quit Google to warn the world about it. The part nobody wanted to hear: > AI is already developing abilities its creators didn't intend > in most cognitive tasks it's already ahead of us > the question is no longer if it surpasses us but when > the only decision left is which side of that line you're on Right now the average person opens Claude, types something, gets an answer, closes the tab. They think they're using AI. they're using maybe 10% of it. I went through his entire lecture, built a practical system from what he was describing. 18 steps to actually use Claude the right way, with copy-paste prompts that work today. Full guide in the post below. Video

x.com/i/article/205463265063…

Claude Code weekly limits are increasing 50%, now through July 13. Live now for all Pro, Max, Team, and seat-based Enterprise users.

Want to (officially) use Codex at work? Send this post to your CTO to bring your team to Codex. Eligible enterprise customers who switch in the next 30 days get 2 free months of Codex usage for new users.

说个暴论，90%的Claude中转站，都在偷偷给你跑Sonnet冒充Opus！ 兄弟们，Claude 中转站里，终于出了一个自己人做的了！ 老板是我朋友，他自己就是重度 Claude Code 用户， 饱受封号之苦，外面的站也用烂了，干脆自己搭了一套，叫 ccode.dev， 号池、防封、风控全链路自研。 用中转站的人，最怕的就一件事就是： 不知道对面给你跑的，到底是啥模型。 所以搭完我第一件事就是拿去测， 一口气跑了两个独立检测平台，结果我直接懵了 ： 1️⃣cctest.ai 跑出来 100%，判定结论直接写着：100%，判定为官方满血 Claude Opus / Max 行为指纹。 2️⃣hvoy.ai 跑了两次，100%，九项全过，延迟 3.5s，速度 17.3 tokens / 秒。 卧槽，那种 "不知道对面是谁" 的感觉，两份报告直接给我打没了。 然后价格方面也是我见过最直接的，100 元人民币，到账 150 美金额度，1:1.5 倍率，真的太香了！ 最反直觉的地方来了，现在公测期充 100 元，正式上线时还额外返 75 额度（按 150 / 2 = 75） 等于现在进去的人，相当于拿内测价，提前享受正式服务。 另外渠道方面也做了分层： 1️⃣官方直连 Max 号池（最稳最快，和官方体验完全一致）， 2️⃣AWS Bedrock 兜底（亚马逊官方渠道，永远不会炸）， 3️⃣逆向渠道备用（极端情况救急，多一层保险）， 大家用的时候按预算和稳定性自己选，别乱冲。 还有个大部分中转站都解决不了的最大痛点， 这次也被解决了—— 以前找中转站，要么贵，要么不稳， 最怕的是出了问题，连个人影都找不到。 但这个站不一样， 假一赔三， 支持开票， 对公也能走。 这一点真的太顶了，最起码出了问题，咱们知道找谁。 如果有需要稳定 Claude 额度的，去 ccode.dev 试试啊~ 遇到啥解决不了的问题也可以直接找我！ Video

GPT-5.6 arriving *that quick* was not on my bingo card.

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: - Claude Agent SDK - claude -p - Claude Code GitHub Actions - Third-party apps built on the Agent SDK

Claude Code weekly limits are increasing 50%, now through July 13. Live now for all Pro, Max, Team, and seat-based Enterprise users.

codex is the best AI coding product and we want to make it easy to try. for the next 30 days, we are giving companies that want to try switching over two months of free codex usage.