2026-07-10

I’m using Opencode and a computer with 128gb. So maybe the results would be different on system. I’ve exhaustingly tried Qwen3.6 27B and Qwen3.6 33B. I have no idea why but they just fall apart when doing more complex tasks with many tool calls. They’re pretty aggressive, doing slightly more than asked, and end up digging themselves into problems. Gemma4 31B and the 26B are literally the opposite. They can’t simply get things done. I have to sit there babysitting them just saying ok, ok, ok. Too

<p> Prompt injection and token savings - #1 in benchmarks </p> <p> <a href="https://www.producthunt.com/products/gate-ai-2?utm_campaign=producthunt-atom-posts-feed&amp;utm_medium=rss-feed&amp;utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1191282?app_id=339">Link</a> </p>

TLDR: 75B-total / 9B-active MoE is the perfect shape for multi-24GB rigs, and almost nobody ships it. Qwen 27B is a great model and punches way above its weight-class, it is a frequent fallback for me. Nemotron-3-Puzzle-75B-A9B, NVFP4, vLLM 0.22.1 (the new Marlin fallbacks run FP4 on Ampere), pipeline-parallel across 3×3090 capped at 200W each. The 4th card runs a speech sidecar untouched - 3 seats × 256K ctx, fp8 KV — hybrid Mamba keeps the cache tiny - 132 t/s decode across 3 streams (~65 sing

I came across this on Hacker News and felt like I needed to share it with the dev community. The main point is simple: a lot of teams reach for extra databases, queues, search engines, caches, and services before they actually need them. This page lays out where Postgres is usually enough, and where you may actually need something else: https://postgresisenough.dev/ Postgres is not perfect for everything, but it is good enough for a surprising amount of real-world work. The more I build and main

Update (07/09/2026): Just pushed a performance optimization. Updating a single shared module improved performance across 13 models by up to 40%! The released ASR models get a 10%+ performance boost. Check https://github.com/0xShug0/audio.cpp/blob/release-0.2/docs/depthwise_conv1d_performance.md I just pushed a new audio.cpp update with streaming support and 4 ASR/STT models: Nemotron 3.5 ASR, Higgs Audio STT, VibeVoice ASR, and Hviske ASR (da only). Overall 1.07x to 2.41x faster than Python. I d

The feud between NightmareEclipse and Microsoft shows no signs of resolving soon.

"Exactly what that dialog looked like between the government and Anthropic and OpenAI is unclear."

“已经出现泛化涌现”

arXiv:2607.07229v1 Announce Type: new Abstract: Prior work has shown that chain-of-thought (CoT) reasoning is often unfaithful: a model's stated reasoning does not reliably reflect the process that produced its output. Detecting unfaithfulness, though, requires controlled experimental interventions, which cannot be applied to evaluation transcripts after the fact. We turn instead to a more tractable question that has received less attention: whether the stated reasoning is logically consistent

arXiv:2607.07097v1 Announce Type: new Abstract: Safety evaluations of multi-agent LLM systems often compare a direct prompt with a planner-executor pipeline and report the difference as a single "pipeline effect." We argue that this aggregate is difficult to interpret because it conflates three mechanisms: harmful intent may be reframed as plausible operational work, the planner may refuse or transform the request, and the executor may act under delegation prompts implying prior approval. To se

arXiv:2607.06906v1 Announce Type: new Abstract: Agentic AI development today runs on token maxing: buying capability with tokens -- longer reasoning traces, more turns, wider tool payloads, bigger replayed contexts -- so tokens per task grow faster than task value. Falling per-token prices mask the pattern; total spend rises anyway. We argue the decisive lever against token maxing is the harness: the orchestration layer that assembles context, exposes tools, sequences turns, delegates work, and

<p> A Computer-Use-Agent that runs legacy software like a human </p> <p> <a href="https://www.producthunt.com/products/coasty?utm_campaign=producthunt-atom-posts-feed&amp;utm_medium=rss-feed&amp;utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1191411?app_id=339">Link</a> </p>

Hey Everyone, here is an update on MTPLX! One month after releasing MTPLX V1 which brought a swift based app and upgraded CLI for coding use I am happy to announce MTPLX V2. The biggest change is Turbo Mode: using custom verify-specialized quantized-matmul kernels plus a compiled verify step we have achieved 82 TPS on a Macbook pro m5 max at a temperature of 0.6 We also released significant changes to SSD KV cache and long context tool calling improvements. here are the preliminary benchmarks fr

https://preview.redd.it/v0xtn3jdu9ch1.png?width=2047&format=png&auto=webp&s=628a6a541fe5f097d0f771ae0ba3b7f44126198f https://preview.redd.it/vjxiucsdu9ch1.png?width=2047&format=png&auto=webp&s=74f7a18a5a30276e206e2bfb5a0c529826ce86e4 This post was originally written in Korean, then polished and translated into English using ChatGPT. I do run llama.cpp locally on a Tesla P40, but as someone who already pays for ChatGPT Pro, I was gradually losing the practical reason to keep running local LLMs li

MOSS-Transcribe-Diarize 0.9B is an end-to-end audio understanding model for long-form multi-speaker transcription, diarization, timestamps, and acoustic event awareness. Given an audio or video file, the model generates a compact speaker-aware transcript in one pass, including timestamps and anonymous speaker labels such as [S01] , [S02] , and beyond. Introduction MOSS-Transcribe-Diarize 0.9B turns real-world long-form audio into structured, speaker-aware transcripts in one pass. Instead of stit

The AI ROI debate has returned and the numbers are even bigger, as are, perhaps, the consequences.

The company is using the cash to open an office in the Bay Area and compete for talent there, "strengthening its position at the heart of the world's leading AI ecosystem."

I don't know where this is headed, but I don't like it. https://futurism.com/artificial-intelligence/open-source-ai-model-scary-mythos GLM-5.2 can be downloaded by anybody, can be run on virtually any hardware, and unlike Mythos or Fable, there’s no vendor playing the middle man between the AI models and the users, raising the cybersecurity stakes considerably. Put simply, while these frontier models can aid researchers in patching holes in commonly used software, the can also be abused by hacke

arXiv:2607.06764v1 Announce Type: new Abstract: Recent progress on ARC-AGI-1 from disclosed architectures has come broadly from two regimes: heavy test-time compute over frontier models (evolutionary search, exhaustive sampling, extended chain-of-thought), or benchmark-specific training in which small models are fine-tuned on ARC data, often with task-specialized architectures. We study a third regime: an open-weight model in non-thinking mode (DeepSeek V3.2) under a strict budget, with no ARC-

arXiv:2607.06720v1 Announce Type: new Abstract: Training large language models (LLMs) with extended reasoning has enabled in-context search, in which models iteratively generate, critique, and revise solution attempts. We provide a theoretical analysis of in-context search by modeling it as approximate inference over reasoning traces, where the base model defines a prior and self-reflection provides feedback for posterior updates, and study the resulting inference-time sampling complexity - the

<p> Full-duplex voice for ChatGPT </p> <p> <a href="https://www.producthunt.com/products/openai?utm_campaign=producthunt-atom-posts-feed&amp;utm_medium=rss-feed&amp;utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1191675?app_id=339">Link</a> </p>

https://artificialanalysis.ai/evaluations/artificial-analysis-openness-index In case you want to support openness, some models are more open than others. Update: K2 think v2 is rated highest because it supplies its training data and training regimen. This allows anyone with enough resources to recreate the model. Deep seek doesn't publish how it trained its model or the training data, so it gets a lower score. If we try to compare software to LLMs. One level of software is that they supply the b

We've been building Stratos and just open-sourced it under AGPL-3.0. It's the layer that turns a raw OpenStack cloud into something you can hand to users and actually bill for. The gap it fills OpenStack already meters usage (Ceilometer/Gnocchi) and can even rate it (CloudKitty). But CloudKitty deliberately stops at "what would this cost" - no invoices, no payments, no taxes, no promotions, no customer-facing portal. So most operators bolt their own thing on top or pay for a closed-source produc

I am curious about intended sizings of the main size niches of the popular local LLM models. As in, we can see there is a major niche at 26b-35b, then hardly anything from 36 through 69b, then (formerly) another major niche at ~70b-72b, then another niche at ~120b-123b, then another big gap till ~230b-235b, and then it gets a bit more mixed all over the place after that with 300b-750b being scattered more randomly probably based more on just whatever the best strength per size they could get whe

OpenAI's No. 2 executive, Fidji Simo, is stepping down from her full-time role after her medical leave proved longer than expected — a leadership vacuum that comes at a tricky time as the company eyes a possible IPO and races to catch Anthropic in the enterprise market.

OpenAI is already shutting down ChatGPT Atlas, its browser that could do tasks for you on your behalf, less than a year after launching it. Atlas was announced in October, but as part of its wave of news about ChatGPT Work today, the company confirmed that it will be "sunsetting" Atlas and is targeting an [&#8230;]

Broadcom accuses Allstate of dodging VMware audits.

Windows 11 updates could soon include fixes for more security issues at once. Microsoft said in a blog post on Thursday that it's now using AI to "identify potential issues earlier," which means "customers will see a higher volume of security updates included in each security release." Hackers, even amateurs, have increasingly been using AI [&#8230;]

arXiv:2607.07021v1 Announce Type: new Abstract: Humans continuously coordinate with others in dynamic interactions, often through implicit, hard-to-quantify social norms that act as shared tacit expectations among interacting agents. As AI agents, including large language models (LLMs), become embedded in daily life, they increasingly participate in such interactions and reshape social interaction structures. Yet they often fail to coordinate with humans in an effective, considerate, and natura

Hi Hacker News, I’m Yahia. I built Context.dev (<a href="https:&#x2F;&#x2F;www.context.dev&#x2F;">https:&#x2F;&#x2F;www.context.dev&#x2F;</a>) to make it really easy to integrate web data into your products and agents.<p>Here’s a demo video: <a href="https:&#x2F;&#x2F;www.tella.tv&#x2F;video&#x2F;build-faster-with-context-dev-apis-2cgl" rel="nofollow">https:&#x2F;&#x2F;www.tella.tv&#x2F;video&#x2F;build-faster-with-context-dev-api...</a><p>Since it’s an API, here are the docs: <a href="https:&#x

AI cheating leads to "a failed society," professor says.

<article><div class="page-wrapper PostDetail-module-scss-module__UQuRMa__hero"><div class="PostDetail-module-scss-module__UQuRMa__illustrationHeroWrapper"><img alt="Inviting hard questions" class="PostDetail-module-scss-module__UQuRMa__heroImage" height="1125" src="https://www-cdn.anthropic.com/images/4zrzovbb/website/3fe58a56e696628496e95871e00b8287035ea645-2000x1125.jpg" width="2000" /></div></div><div class="page-wrapper"><article><div class=""><div class="Body-module-scss-module__z40yvW__bod

Hello everyone! Pangolin 1.20 is focused on how people find and reach their resources in the UI. Here's what's new: Pangolin is an open-source, identity-based remote access platform that lets you securely expose your infrastructure to your team. It supports browser based remote access and a remote access VPN in one platform with strong authentication controls. GitHub: https://github.com/fosrl/pangolin Resource Launcher The landing page non-admins see when they sign in now is vastly more capable.