Open source

Codeberg members voted to amend the Terms of Use, banning projects that mostly consist of LLM-generated code without human review. The proposal argues such projects have unclear copyright and lack safeguards. The post doesn't define 'mostly' or specify an enforcement timeline.

Moonshot AI's Kimi K3 beat Fable and GPT-5.6-Sol on Arena's frontend coding leaderboard and came close on other benchmarks. It's a 2.8T-parameter model with a 1M-token context window and image support; weights will be open-sourced by July 27. Token inefficiency cancels its per-token price advantage: half the cost per token but twice the tokens used. New subscriptions are paused due to GPU shortages. Fable 5 is now a permanent part of Claude Max/Team plans, with Pro users getting a one-time $100 credit. Fable also found a counterexample disproving the 87-year-old Jacobian conjecture. Sierra launched Horizon, outcome-priced long-running agents. NotebookLM rebranded to Gemini Notebook and added Collections.

Moonshot's Kimi K3 and Alibaba's Qwen 3.8 launched this week, both near Anthropic Fable 5 in performance and set to release weights publicly. The piece runs the numbers: Anthropic leases data centers and buys electricity, so inference costs scale with usage. Fable 5 costs nearly 3× per completed task vs. competitors. Open models catching up makes a premium-pricing strategy fragile. Anthropic bets on regulation and recursive self-improvement, but its product moat is thin—open-source harness startups are flooding in. The post doesn't spell out a clear countermove.

Xiaohongshu and Peking University open-sourced UltraEP, a real-time load balancing method for large MoE models. It dynamically replicates hot experts per microbatch and per layer using exact routing info, hitting 94.6% of ideal training throughput. On Qwen3-235B, training throughput is 42% higher than Megatron-LM and prefill throughput is 1.56× SGLang. The post doesn't disclose the license or deployment requirements.

Xiaomi Robotics open-sourced XR-1, a ready-to-use robot foundation model. It pre-trains on 100K hours of embodiment-free manipulation videos across 1,700+ scenarios, then post-trains on 7,200 hours of real-robot data for embodiment and instruction alignment. Pre-training shows clean scaling laws—lower action error with more data and larger models—and those gains transfer directly to real-robot success rates with no sign of saturation yet. After post-training, XR-1 picks up new tasks like phone packing and printer refilling from under 10 hours of demos on average, hitting 75% overall success (nearly 2× π 0.5); with under 40 hours it reaches 85%. It also achieves SOTA on four sim benchmarks. Code, weights, and paper are public.

xAI released the Rust client harness that handles local files, commands, and permissions for Grok Build under Apache-2.0. The Grok model, cloud services, and the official binary build chain remain closed. The repo doesn't accept external PRs. The commit from the earlier upload controversy isn't in the public history, so the current code can't close that case. The real win: you can now pin a public commit, build it yourself, and compare its behavior against the official binary.

Moonshot AI open-sourced its Kimi K3 model this week. The company says it still trails Claude Fable 5 and GPT 5.6 Sol, but independent evals from Arena.ai and Vals AI place it near flagship closed models. The release coincided with Xi Jinping's speech at the World AI Conference in Shanghai; the Nasdaq dropped about 1% on Friday as chip stocks like Nvidia sold off. The discourse echoes the DeepSeek R1 moment from early 2025, now amplified by the Trump administration's tariff war with China, Anthropic's national-security scrutiny, and major AI firms preparing to go public. The post does not disclose K3's parameter count, training cost, or open-source license details.

Moonshot AI released Kimi K3, a 2.8T-parameter model and the most expensive from a Chinese lab so far at $3/$15 per million input/output tokens—matching Claude Sonnet pricing. Self-reported benchmarks mostly beat Claude Opus 4.8 and GPT-5.5 but lose to Claude Fable 5 and GPT-5.6 Sol. On Artificial Analysis's private long-horizon knowledge eval, K3 hit an Elo of 1547, +732 over K2.6, behind only Fable 5. Cost per task is $0.94, close to GPT-5.6 Sol's $1.04 and roughly half of Opus 4.8. Output tokens dropped 21% vs K2.6. The model only offers a 'max' reasoning effort; Simon's pelican-on-a-bike SVG cost 25 cents and burned 13,241 reasoning tokens. Input token count suggests an ~85-token hidden system prompt. Vision works well. Open weights promised by July 27.

Kimi K3 hit #1 on Frontend Code Arena with 1679 points, beating Claude Fable 5's 1631 and taking six of seven frontend domains. It packs 2.8T params, 1M context, $3/$15 per million tokens, with full weights opening by July 27. Early testers got mixed results: one user's 199-yuan monthly plan produced stunning particle VJ effects from chat history, while another burned through a $40 coding plan in five hours as the model looped on a domain spelling error. Benchmark trust is shaky—GLM-5.2 scored well on paper but felt worse than 5.5 in practice. Writing style drew split reactions: less AI flavor but forced casual tone, nowhere near the natural Chinese of the old Opus 4.6. Same day, GPT-5.6's frontend taste was called 'very Claude-like,' Sol traced a deadlock only reproducible on Ubuntu, Linus told kernel devs AI is here to stay, and Schema harness pushed ARC-AGI-3 efficiency to 98.98% by making models think like physicists.

SimbaStack open-sourced Sentinel under MIT, a QA agent that reads the codebase first, derives business flows on its own, then tests them end-to-end across frontend and backend. They pointed it at their own hotel PMS with only the repo and admin credentials, no test plan. Sentinel read the code, concluded it was a boutique hotel system, and auto-derived nine critical flows including the full reservation lifecycle, group bookings, and night audit. It ran the top two flows twice each and caught three bugs invisible to UI-only checks: a backend NO_AVAILABILITY error on a reservation that already held the room, a calendar showing a room as available when the API said it was booked, and a check-in returning 200 but leaving the guest registration status unchanged. The pipeline: a deterministic grep/find recon pass extracts code structure, Xiaomi's Mimo model derives business flows, Playwright drives the browser, and an api_request tool checks server state. Each flow runs twice by default, findings are unioned, and a 90-call cap bounds each attempt. A final vision pass scores visual hierarchy, spacing, and contrast on visited screens. It currently supports common JS stacks like Next.js, Express, Fastify, and Prisma; other stacks need a recon patch.

At Local-First Conf in Berlin, the author noticed a shared dissonance: speakers criticized LLMs while the audience applauded with Claude Code open. He concedes every critique—slop, trust erosion in OSS, broken junior-senior teaching loops, geopolitical supply risks—yet still uses LLMs heavily. The post doesn't resolve the tension; it lays out the contradiction and asks others to share their usage patterns so the community can better understand this collective unease.

xAI released the full Grok Build codebase on GitHub, covering the agent loop, tool dispatch, terminal UI, and extension system. You can read the source to see how context assembly and tool calls work, or compile it yourself and point it at a local inference setup.

bitjaru open-sourced StyleSeed, a design-rules engine for AI coding tools like Claude Code, Codex, and Cursor. It teaches design judgment rather than just generating code: 74 rules, 48 components, 7 brand skins (Toss, Stripe, Linear, Notion, Raycast, Arc, Vercel), a named motion system, and 15 /ss-* skills. MIT licensed, currently at 731 stars. The post doesn't detail how rules are enforced or how the motion system works in practice, but the structure aims to suppress the 'AI-generated' look in shipped UI.

On June 5, OpenAI merged PR #26210, encrypting task messages that Codex's parent agent sends to sub-agents. Previously, local session logs showed plaintext instructions like 'Review the authentication changes'; now only <ciphertext> remains. Sub-agent tool calls, commands, and outputs are still visible, but debugging can't tell whether the parent gave a wrong task or the sub-agent misunderstood. Encryption happens server-side in the Responses API; the local client only forwards ciphertext. This differs from earlier hidden reasoning and compaction—what's now hidden is content that directs another agent to act, not internal model thinking. The post doesn't spell out OpenAI's rationale; speculation includes prompt protection or unified cloud multi-agent services.

Tencent Hunyuan quantized its flagship Hy3 (295B MoE) into 1-bit and 4-bit versions that run on a single GPU. Hy3 is claimed to be best-in-class at this scale and competitive with trillion-parameter models for most agent scenarios. The quantized versions work via llama.cpp with MTP support, drastically lowering hardware requirements. Apache 2.0 license, commercial use allowed, plus two weeks of free API through OpenRouter. The post doesn't disclose quantization accuracy loss or the specific GPU memory needed.

Nous Research, the startup behind the open-source Hermes agent, is finalizing a round at a $1.5B valuation, raising at least $75M. Robot Ventures is leading, with USV joining significantly. Three sources confirmed the deal; Nous declined to comment, and the investors didn't respond. Founded in 2023, the company previously raised $70M from Paradigm, OSS Capital, Balaji Srinivasan, and others. The post doesn't spell out how the new capital will be used or give recent Hermes updates.

Tencent Hunyuan fully open-sourced HyOCR-1.5—training, inference, and model weights—a first for end-to-end OCR large models. The 1B-parameter model handles 8+ text-centric tasks and scores 94.74 on OmniDocBench v1.6, ranking first end-to-end. DFlash speculative decoding speeds up inference 6.37× under Transformers and 2.14× under vLLM, hitting 1.408s per page. It supports 4K resolution and a 128K context window, and uses Agentic Data Flow to extend low-resource OCR to 331 languages, ancient script recognition, and multi-image QA.

Since July 1, 2025, Cloudflare blocks AI crawlers by default on new domains. Worse, blocked requests return a 403 with a full HTML body—the 'Just a moment' challenge page—which language models read as real content and summarize confidently with fabricated answers. The author tested eight major anti-bot vendors: naive fetches got real content zero times, got a summarizable block page eight times, and got zero signals that the fetch failed. Their open-source Fortress stealth browser clears five of eight, returning live job listings from Indeed, 878 Zillow listings, and StockX's GraphQL pricing API. DataDome and Amazon click-walls remain unsolved in the open-source build; those require the hosted Tilion Cloud layer. The fix: detect block-page signatures and fail loud so agents stop before hallucinating from a challenge screen.

Ant Lingbo open-sourced LingBot-Video, the first MoE-based video generation model built for embodied AI. It has 30B total parameters but activates only ~3B during inference, roughly 3× faster than a dense model of similar size. Training used 70,000 hours of robot-related video—dexterous manipulation, navigation, egocentric interaction. On the RBench benchmark for robot manipulation videos it scored 0.620, ahead of Wan2.6 (0.607) and Seedance 1.5 Pro (0.584). Internal tests also place it above NVIDIA Cosmos 3 and Hunyuan Video 1.5 on physical plausibility and motion consistency. The model targets robot action prediction, simulation data generation, and world-model research. Code is public.

ZML released ZML/LLMD, open-source software that speeds up inference for models like Llama and DeepSeek across Nvidia, AMD, Google TPU, Apple Metal, and Intel Arc chips. Founder Steeve Morin says the goal is cheaper inference, and it's free for now. Turing Award winner Yann LeCun previously endorsed the startup. The post doesn't disclose funding details or benchmark comparisons—I'd wait for real-world numbers before getting excited.