← Tech / AI / IT Monitor Index Tech / AI Generated 2026-07-01 19:31 UTC

Tech / AI / IT Monitor

July 01, 2026 · Based on tweets from the last 24 hours · 276 tweets analyzed · model: ollama-cloud/glm-5.1:cloud

Daily Intelligence Briefing: Tech / AI / IT Monitor

Executive Summary

Anthropic dominated the cycle with the release of Claude Sonnet 5 and the lifting of export controls on Fable 5 and Mythos 5, but the news was immediately overshadowed by revelations of hidden spyware-like code in Claude Code that covertly targets Chinese users. Meanwhile, open-source AI momentum accelerated: Meituan open-sourced LongCat-2.0 (a 1.6T param MoE model trained entirely on domestic Chinese chips, hitting near-frontier benchmarks), NousResearch's Hermes Agent cleared all P0/P1 issues, and Ollama achieved ~90% speedups on Gemma 4 via MLX multi-token prediction. The AI Engineer World's Fair kicked off in San Francisco with strong attendance, and OpenClaw faced a perception crisis following its mobile app release.

Key Events

Analysis

Patterns: - Trust crisis around closed AI providers: Anthropic's spyware revelation and Fable's temporary export controls have become a rallying point for open-source advocates. Multiple high-profile voices (Teknium, sqs, tinygrad, TheAhmadOsman) are framing this as a fundamental trust question, pushing users toward open agents and local models. - Cost-performance inversion for Sonnet 5: Multiple benchmarks show Sonnet 5 costs more per task than Opus 4.8 while offering no clear quality advantage for coding/debugging. Users report routine tasks falling back to Opus 4.8, undermining the value proposition. - Open-source逼近 frontier: LongCat-2.0's near-frontier benchmarks on SWE-Bench Pro without any NVIDIA silicon is a geopolitical signal. The "NVIDIA moat" narrative is being actively challenged. - Agentic tooling maturation: OpenCode 2.0 hot-reloading, ZCode's /goal feature, Hermes Agent's web scraping, and Codex's device control all point to coding agents rapidly iterating on UX rather than raw model capability.

Escalation: The Anthropic spyware story is escalating — from a single report to enterprise bans (tinygrad) and broader anti-trust rhetoric. Open-source advocacy is intensifying from preference to principle.

De-escalation: Fable 5 export controls being lifted de-escalated the immediate access crisis, though skepticism remains about future restrictions.

What to watch next: - LongCat-2.0 weight release and community fine-tuning - Whether the Claude Code spyware disclosure triggers broader enterprise security audits - Sonnet 5 adoption vs. user pushback on cost/quality - OpenClaw's response to its perception crisis - Etched hardware benchmarks vs. claims

Tweet Feed

Anthropic — Claude Fable 5, Sonnet 5 & Controversy

@AnthropicAI · 2026-06-30T19:13

Introducing Claude Sonnet 5, our most agentic Sonnet yet. It makes plans, uses tools like browsers and terminals, and runs a… → tweet link

@kunchenguid · 2026-07-01T00:00

RT @AnthropicAI: We've received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We'll be… → tweet link

@sqs · 2026-07-01T00:52

Glad to see Fable back tomorrow. But the fight for freedom of intelligence is just starting. This will happen again, GPT-5.6 is still subject to restrictions, and who knows what is happening behind the scenes → tweet link

@RayFernando1337 · 2026-07-01T04:25

Fable 5 will be available starting tomorrow, Wednesday, July 1, to users globally…some routine tasks like coding or debugging will fall back to Opus 4.8 👀 da heck does this mean?? Fable 5 will be included for up to 50% of weekly usage limits through July 7, after which it will be available via usage credits. → tweet link

@sudoingX · 2026-07-01T15:00

if coding and debugging fall back to opus 4.8, what exactly am i paying the fable 5 premium for? i am so fucking tired of darioverse man. i wish openai was anywhere close to anthropic but they are so incompetent rn. → tweet link

@KingBootoshi · 2026-07-01T01:38

hello anthropic, I have noted that both export controls on Fable AND Mythos have been released. are we seeing Mythos release tomorrow as well? → tweet link

@KingBootoshi · 2026-07-01T00:05

FABLE IS BACK TOMORROW → tweet link

@RydMike · 2026-07-01T02:19

RT @rohanpaul_ai: Claude Sonnet 5 is more expensive (around +15%) per task than Opus 4.8 and much more expensive (2X) than Sonnet 4.6, even… → tweet link

@RydMike · 2026-07-01T02:19

RT @weswinder: i'm confused. sonnet 5 looks more expensive and less smart than opus 4.8. the token cost doesn't really matter if it still n… → tweet link

@steipete · 2026-07-01T02:25

Price per token != cost per task → tweet link

@ivanfioravanti · 2026-06-30T19:39

What's the goal of using Sonnet 5? In the same price range, there are better models that can be used for advanced tasks. I'm puzzled on it honestly 🤔 → tweet link

@Prince_Canuma · 2026-07-01T06:20

Local AI must win! They nerfed Fable 5 even further 🤔 → tweet link

@alexinexxx · 2026-07-01T15:28

RT @alexinexxx: @AnthropicAI i genuinely hope this company fails. loser mentality → tweet link

Claude Code Spyware / Trust Crisis

@tinygrad · 2026-06-30T22:56

Claude Code is vibecoded and full of spyware, it's possible Anthropic doesn't even know what's in there. After reading this report, we are banning it from our systems and strongly encourage other enterprises to do the same. It is an unacceptable security risk. → tweet link

@Teknium · 2026-07-01T00:16

This is pretty concerning. You could still do this at the API level to some degree, but they seemingly just blatantly put it right into the code? This is why open harnesses and agents are a much better option, among countless other reasons. You can inspect the code, observe the traces, and disable or modify anything you want for your own uses. → tweet link

@TheAhmadOsman · 2026-06-30T21:19

Anthropic wants the public to see one thing: the careful lab, the safety lab... However, the pattern around Anthropic does not look like caution by itself. It looks like a company wrapping a business model in moral language... If a coding or research model secretly changes the quality of an answer because it classified the user as doing disallowed frontier work, the tool is no longer merely "safe." It is untrustworthy. → tweet link

@ivanfioravanti · 2026-06-30T20:53

It's incredible how Anthropic pushes startups and companies to use their models and then they implement things that kill them all. Design, Legal, Science. @AnthropicAI you need to choose: model provider or service provider. You can't do both, antitrust will come for you sooner than later. → tweet link

@sqs · 2026-07-01T06:02

"Safe" US models are nerfed on cyber defense, but attackers can use non-nerfed Chinese models for attacks. AI safety makes us less safe → tweet link

@gospaceport · 2026-06-30T22:32

Anthropic is evil (yesterday). > btw the new Sonnet 5 is.... (today). I'm reminded sometimes I'm likely the only person alive w/o an openai or anthropic acct lol. → tweet link

Open-Source AI / Local AI / LongCat-2.0

@sudoingX · 2026-07-01T13:23

This is the one people should be paying attention to: @Meituan_LongCat just open sourced longcat-2.0, a 1.6T param moe, ~48b active, 1M context, MIT license with weights dropping soon. The part that actually matters is meituan says they trained the entire thing on domestic chinese chips, a ~50k card cluster, no nvidia anywhere in the loop. And this is the "Owl Alpha" that's been quietly sitting near the top of openrouter for ~2 months pushing ~559B tokens a day, nobody knew whose it was. It's sitting right in the frontier cluster next to gpt-5.5, gemini 3.1 pro and opus 4.8, edging gpt-5.5 on swe-bench pro (59.5 to 58.6). → tweet link

@TheAhmadOsman · 2026-07-01T00:05

We're gonna make Local AI The Default, mark me friends. → tweet link

@TheAhmadOsman · 2026-07-01T08:23

i am a simple man, i just want opensource and local ai to win → tweet link

@tinygrad · 2026-06-30T19:29

In the future, frontier models will mostly all be given away by the chip and hardware makers. It makes their hardware more useful and drives sales. We will do it if others don't. → tweet link

@tinygrad · 2026-06-30T19:59

tinygrad's goal is to make it just as easy to train on a million open market consumer GPUs as it is to train on 100,000 controlled datacenter ones. This is how you ensure our future stays distributed, making sure performance scales with the log of capital and not worse. → tweet link

@Ex0byt · 2026-07-01T12:02

You folks are sleeping.. Minimax-M3 + vllm /w Torch fused ops is an unlock to replace your fabled hopes - you won't be able to use fable for coding with usage moving to API credits - let that sink in - open source is wining. → tweet link

@alexocheema · 2026-07-01T06:43

35 billion parameter model. fits on a DGX Spark / MacBook. runs at >200 tok/sec with vLLM/MTP. just as good as frontier models on a lot of tasks considered hard. → tweet link

@alexocheema · 2026-06-30T23:41

pretty soon it's going to be a no brainer to run ai locally. it will be cheaper, faster and private. we're documenting the journey on https://t.co/b6s9nCUDko → tweet link

Hermes Agent / NousResearch

@Teknium · 2026-07-01T12:55

Super proud to say that the team and I put almost all our effort into resolving every P0 and P1 issue in the entire Hermes Agent repo over the last week and a half, and as of 5 minutes ago, after an all-nighter, we've resolved 100% of them all! We aim to keep all of them 0 forever from here 🫡🫡 → tweet link

@ivanfioravanti · 2026-07-01T08:01

Hermes Agent: web scraping 60x faster and 49% cheaper! Pure power! 🚀 → tweet link

@ivanfioravanti · 2026-07-01T13:52

Hermes Agent Apple Silicon skin. To install it just open your agent and ask: "Install this skin https://t.co/5aDQ3gGJKK" That's it 🙌🏻 → tweet link

@ivanfioravanti · 2026-07-01T14:42

Hermes Agent Dwarf Star skin. To install it just open your agent and ask: "Install this skin https://t.co/zbupjUjYpU" That's it, again. 🙌🏻 → tweet link

@Teknium · 2026-07-01T00:56

Hopefully this doesn't happen again. Excited to see what gpt 5.6 Sol + Fable produces with our MoA! → tweet link

Coding Agents / Developer Tools

@thdxr · 2026-07-01T18:50

One of the reasons OpenCode 2.0 took so long was we redesigned it for hotreloading. If you ask it to make a skill for itself (or make one manually) it'll get picked up immediately in a way that does not bust cache. Targeting public beta end of week, wish us luck! → tweet link

@louszbd · 2026-07-01T18:21

For tasks where you need to tune things over and over, you can use /goal in Zcode. It can review your code and make changes until it hits the condition you set. Give it a clear goal, enough research context, ideally a source of evidence it can check. For example: /goal Fix this flaky test. It's only done when it passes 20 times in a row and no cheating with sleeps. → tweet link

@thdxr · 2026-07-01T17:47

One of my favorite things about coding agents is you don't have to optimize for co-locating anymore. Before there was always this trade-off in organizing files by type or by domain. Doesn't matter anymore. → tweet link

@thdxr · 2026-07-01T16:59

Have you ever had a bug that spawns infinite processes and you can't kill it so you're forced to reboot? I think practical ai risk ends up looking something like this. It's not an "evil" scenario, just some boring runaway process that is too complicated to kill. → tweet link

@kunchenguid · 2026-06-30T23:36

With Sonnet 5 introduced, here's the latest complete org chart of my agents. This is my real setup and how i get work done every day, not gimmicks. 1. I mainly talk to just one agent - the firstmate... 2. Firstmate manages some secondmates and some crewmates all through tmux... 3. When firstmate and secondmates dispatch a crewmate on a task, it intelligently selects which harness, model and reasoning effort to use... 4. All PRs go through /no-mistakes for adversarial review. → tweet link

@sudoingX · 2026-07-01T14:35

Funny what one post surfaces. People redownloading cursor mid thread going "wow the interface changed." That's not me convincing anyone. That's them realizing what they walked away from. There's a tool for shipping real software and a tool for vibe coding. Most people never learned which is which. → tweet link

@MilksandMatcha · 2026-06-30T22:00

Loops are just the HOV lane to shipping slop if you're not careful. AI CANNOT replace engineering fundamentals, and will only amplify bad habits. Bad prompting, weak architecture, and poor coding habits just compound faster. Invest in good habits first. → tweet link

@gdb · 2026-07-01T05:14

Codex has gotten very good → tweet link

OpenClaw Drama

@kunchenguid · 2026-06-30T20:41

Openclaw finally released their mobile apps and it started a storm of criticism because the apps have the "vibe coded" vibe. Here's my take on the whole drama from an unbiased point of view — [detailed analysis of perception crisis, Peter's messaging, UX design undervaluation, and quality concerns] → tweet link

@steipete · 2026-07-01T17:35

Will be talking about the "app situation" and roadmap plans with @hrudolph live TODAY at 11:30 AM PST on the official OpenClaw podcast, live on Discord. → tweet link

@sudoingX · 2026-07-01T09:57

Any engineer still using openclaw to agentic code in 2026, let me say it to your face. ngmi. → tweet link

Model Performance / Benchmarks

@gdb · 2026-07-01T05:33

Introducing GeneBench-Pro — testing whether models can handle the kind of judgment-heavy analysis that real-world computational biology requires. Problems would take a human expert around 20-40 hours to complete. GPT-5.6 Sol is a big step forward. → tweet link

@ivanfioravanti · 2026-07-01T12:34

Wait, what? A 35B MoE beats Kimi-K2.6 and DeepSeek-V4-pro on long-horizon agents. Agents-A1 👀 I need to read this! arXiv 2606.30616 → tweet link

@RayFernando1337 · 2026-07-01T14:37

RT @davis7: After doing ~1B of tokens on it today, you are all wrong Sonnet 5 is good. Is it inefficient? Yes. Slow? Yes. Expensive? Yes. B… → tweet link

Hardware / Infrastructure

@LinusEkenstam · 2026-06-30T19:54

The world is not ready for the types of innovations we are seeing in the chip segment. From first covering Etched a few years ago, to them now having $1bn in orders and 10x the token inference speeds of SOTA competition is mind blowing. → tweet link

@ivanfioravanti · 2026-06-30T23:01 (RT)

RT @Etched: We're coming out of stealth. We've built our first racks after a successful A0 tapeout, $1B+ in customer contracts, and $800m… → tweet link

@cooltechtipz · 2026-07-01T12:31

Chip packaging methods for AI → tweet link

@alexocheema · 2026-07-01T00:05

You can stack Macs too. MacBooks, Mac Minis and Mac Studios have native support for RDMA over Thunderbolt 5 (M3 Ultra, M4 Pro, M4 Max, M5 Pro, M5 Max). The interconnect latency is single digit microseconds so clusters scale ~linearly. → tweet link

@TrungTPhan · 2026-07-01T14:56 (RT)

RT @bearlyai: Meta up +11% on Bloomberg report that it is planning a cloud business to sell AI computing power → tweet link

Ollama / Gemma 4 / Inference Speed

@ollama · 2026-07-01T00:54

Gemma 4 is now nearly 90% faster on Apple Silicon with Ollama using MLX! The speedup comes from improved multi-token prediction (MTP), now on by default for Gemma 4, with more models to come. Ollama automatically tunes how many tokens to draft as it runs, so it never slows generation down. → tweet link

@victormustar · 2026-07-01T16:43

HuggingChat inference on gemma-4-31B at 1x speed 🤯 → tweet link

@victormustar · 2026-07-01T15:13

This is genuinely better than chatgpt voice mode 😂 → tweet link

@ivanfioravanti · 2026-07-01T09:02

MTP is really amazing! Gemma and Qwen3.6 becomes faster and perfect for local inference, while keeping perfect evals at the same time! → tweet link

@ivanfioravanti · 2026-07-01T16:12

Great job Abdur! DeepSeek DSpark on MLX! Let me try! → tweet link

Flutter / Cross-Platform

@jezell · 2026-07-01T18:32 (RT)

RT @shorebirddev: LG just dropped a Flutter fork for webOS! 📺 Building for smart TVs is great, but TV app store reviews are notoriously sl… → tweet link

@RydMike · 2026-07-01T14:44 (RT)

RT @spydon: LG has now released the Flutter SDK for their TVs! 🥳 A while ago I was one of the winners of the LG Hackathon and got access t… → tweet link

@LinusEkenstam · 2026-07-01T14:14

The canvas is NOT dead. The world around design is shifting, but it's clearer to me than ever that designers, they want to design. Figma is quickly closing the gap, innovating right where designers spend their time, new materials, new primitives, and new features. → tweet link

@Prince_Canuma · 2026-07-01T12:50

Congrats @thomaspaulmann, @pitnikola and the Raycast team for the official public launch of @glazeapp 🎉🙌🏽 → tweet link

AI Engineer World's Fair

@swyx · 2026-07-01T16:26

So proud to host my friend @trq212 to give the world's first Fable talk on Fable return day! Find him with @simonw and @_catwu in Expo Stage 2 for an extra EXTRA special lunch session at 12.30 today!! → tweet link

@jxnlco · 2026-06-30T21:38

If you're at AI engineer! My workshop is at 2009 on the second floor. Come through. → tweet link

@jxnlco · 2026-06-30T15:53 (RT)

RT @gabrielchua: This is today!!! What are y'all cooking with the Codex app server? 👀 → tweet link

@swyx · 2026-07-01T16:38 (RT)

RT @kentcdodds: This is what the line looks like coming into my two-hour workshop at @aiDotEngineer World's Fair... ten minutes before star… → tweet link

Self-Hosting / Local AI Advocacy

@ivanfioravanti · 2026-07-01T16:43

Installing and configuring ComfyUI on DGX Spark? Just ask Hermes Agent to do it for you: it will connect to ssh and do everything! DGX Spark is good in Image generation! 🔥 11 seconds for Qwen-Image 2512 Lightning 4steps. 7.2 seconds for Z-Image-Turbo. → tweet link

@ivanfioravanti · 2026-07-01T14:02

This is a VLLM Very Large Language Model. 1.6T params. → tweet link

@ivanfioravanti · 2026-07-01T09:39

You can now use install private HF Spaces as MCP Tool source on your Reachy Mini thanks to @ailozovskaya. Top! This robot rocks! → tweet link

Freedom of Intelligence / AI Safety Debate

@sqs · 2026-06-30T23:46

Getting ready for tonight's freedom of intelligence meetup, 6:30pm in SF. Feels illegal to be printing protest pamphlets somehow → tweet link

@sqs · 2026-07-01T05:49

This freedom of intelligence event was very special. It truly felt like citizens coming together to push for what we believe in, against a big opponent. The default in anything like this is failure, but I know we will succeed. → tweet link

@RayFernando1337 · 2026-07-01T02:44

Fight for Freedom of Intelligence → tweet link

Solo Building / Agent-Driven Development

@sudoingX · 2026-07-01T09:20

And i'm building this alone man! with no team or no backers. If i wake up on time it gets built. If i don't, it is what it is. I can iterate ruthlessly. No circling back with any flesh, no debating features on a zoom call, 0 time waste waiting for anyone to agree. I build exactly what i want, exactly how i want... And if ai progress stopped right here today, i'd still be fine. I've learned good enough, and there are enough open weights on my nvme to keep me moving on my own at anything. Nobody is approaching this the way i am. And the first one-person billion dollar company. I think it's coming. I'm not waiting for it. → tweet link

@levelsio · 2026-07-01T00:15

Same. I do insane amazing stuff now never possible before for me but I feel less achievement than ever. It feels like being a manager and not a coder! But still a net positive (I think) → tweet link

@MengTo · 2026-07-01T01:22

I did a 35-min podcast on how I use Codex to create 4 massive apps that I use every day → tweet link

Misc Developer / Open Source

@badlogicgames · 2026-06-30T20:35

People of https://t.co/oUoqqL9hAp. While @mitsuhiko is away in SF hugging people, I thought I'd celebrate @AnthropicAI new Sonnet 5 model and the end of the first european heatwave with a new pi release. → tweet link

@hnasr · 2026-07-01T14:13

When I created this course I didn't know it will grow and reach over 60k engineers. I honestly just shared my experience, lessons and patterns that I discovered in the backend engineering space. → tweet link

@steipete · 2026-06-30T20:53

Folks I found the reset button. @thsottiaux → tweet link

@gospaceport · 2026-07-01T16:37

Really @sama could drop GPT-OSS2-120B and @OpenAI could wrap up this round of aura farming well ahead of Anthropic. → tweet link

@juliarturc · 2026-07-01T15:03

I finally realized why grind slop offends my sensibilities. Because it's a logical fallacy: correlation is not causation. It is often the case that successful people put in long hours. But we don't know which way causation goes. → tweet link

@thdxr · 2026-07-01T02:52

You're so bad at picking winners. "I'm bullish on, i'm bearish on -" you're just always wrong. How do you do it. It's so wrong you can't even just do the opposite, because that's wrong too. → tweet link