Executive Summary
The past 24 hours highlight a strong push towards localized, agentic AI and growing friction around closed-model dependencies. Qwen 3.7's API-only release sparked significant community pushback, with developers demanding open weights to prevent a drift toward closed ecosystems. Simultaneously, practical demonstrations of local AI capabilities continue to impress, as a decade-old GTX 1080 was shown running Qwen3 8B autonomously. In enterprise AI, Microsoft is reportedly forcing internal migration from Claude Code to Copilot for political reasons, while Valve flagged exploding AI API costs. Meanwhile, the agentic ecosystem matures rapidly, with Hermes Agent expanding multi-tool orchestration and xAI's Grok Build gaining traction for its real-time social data integration.
Key Events
- Qwen 3.7 API-only release sparks open-weight backlash: The release of Qwen 3.7 without open weights has caused frustration in the local AI community, with developers urging Alibaba to keep the ecosystem open. → link
- Microsoft migrating employees from Claude Code to Copilot: According to industry analysis, Microsoft is dropping Claude Code licenses not due to cost, but internal politics and dogfooding pressure for Copilot. → link
- Valve reportedly tells engineers to stop using Claude due to cost: A viral tweet highlights that Valve's AI bills are exploding, signaling a broader cost crisis for enterprise API usage. → link
- Local agentic AI runs on 10-year-old GPU: A developer successfully ran Qwen3 8B locally at 18-20 tok/s via Hermes Agent on a GTX 1080, even autonomously bootstrapping a development environment. → link
- PrismML releases 1-bit and Ternary Bonsai Image 4B: A new family of highly compressed image-generation models designed to run high-quality generation efficiently was released. → link
- Hermes Agent expands orchestration capabilities: Nous Research's Hermes Agent now orchestrates OpenHands, Claude Code, Codex, and OpenCode through optional built-in skills. → link
- Grok Build emerges as a contender in CLI coding agents: xAI’s Grok Build is gaining attention for integrating real-time X feeds into a CLI coding agent, bundled with X Premium+. → link
- Google AI Studio adds native Android app generation: Google AI Studio now allows users to build native Android apps. → link
Analysis
Patterns: There is a clear and intensifying tension between closed, expensive API dependencies and the open-source/local inference community. Sentiment is shifting toward local ownership, driven by exorbitant enterprise API costs (Valve, Microsoft) and strategic closed-weights shifts (Qwen 3.7). Meanwhile, AI agents are rapidly consolidating into orchestration hubs—meta-agents like Hermes are unifying disparate coding agents into single workflows.
Escalation/De-escalation: AI API cost crises are escalating within large enterprises, forcing hard reversals on tool usage. This is de-escalating the "rent vs. own" debate: developers are increasingly viewing local GPU ownership as a necessary hedge against unpredictable token billing and nerfed models.
What to watch next: * Whether Alibaba/Qwen yields to community pressure and releases open weights for 3.7. * The adoption curve of Grok Build, specifically whether its real-time X feed integration provides a durable competitive moat in the CLI agent space. * How enterprise cost-cutting measures (like Microsoft's Copilot mandate) affect developer productivity and sentiment.
Tweet Feed
AI Models & Open Source Weights
@sudoingX · 2026-05-26 17:42
local builders kept qwen ecosystem alive across 3.5 and 3.6 open weights releases. that's why qwen is THE name in open weight inference in 2026. don't break the pattern on 3.7.... 3.7 is currently api only. weights still tba. one repost on this thread is one more signal alibaba sees. let's see if the gate moves. → tweet link
@sudoingX · 2026-05-26 10:37
qwen 3.7 api results look insane. when open weights? don't leave us local builders behind. if we want qwen 3.7 open weights, like and repost. let them hear us. → tweet link
@victormustar · 2026-05-26 18:36
RT @PrismML: Today we’re releasing 1-bit and Ternary Bonsai Image 4B. A new family of image-generation models designed to run high-quality… → tweet link
@TheAhmadOsman · 2026-05-26 15:12
MiniMax M3 👀 https://t.co/sb10cSjj8l → tweet link
@TheAhmadOsman · 2026-05-25 23:23
Buy a GPU tried to warn you of this in 280 characters or less https://t.co/TuGPKTjvxJ → tweet link
@TheAhmadOsman · 2026-05-25 20:32
The single highest-leverage move an individual can still make in 2026 is this: Buy a GPU and run real workloads on them every day. - Not rent tokens. - Not fine-tune on someone else’s cluster. - Own the silicon + the weights + the serving stack... → tweet link
@thdxr · 2026-05-25 23:01
there are a lot of reasons to want to run models locally cost is probably not one of them → tweet link
@steipete · 2026-05-26 11:10
RT @dhh: I've had more "I can't believe it's this good" moments with GPT5.5 than any other model since Opus 4.5. It's shockingly, scarily c… → tweet link
@thdxr · 2026-05-26 01:17
this is the best model i've ever used https://t.co/NQpTzfLEv8 → tweet link
AI Agents & Developer Tools
@sudoingX · 2026-05-26 18:53
ok this is wild. 10 year old gtx 1080 8gb pascal card running qwen3 8b locally at 18-20 tok/s via hermes agent and it's actually doing the thing... agent is firing tool calls cleanly and the model is reasoning about next steps at 18-20 tok/s. on hardware that pre dates "agentic" as a word. → tweet link
@Teknium · 2026-05-25 22:28
Hermes Agent now can orchestrate the @OpenHandsDev agents with a new optional skill!
hermes updatethen dohermes skills install official/autonomous-ai-agents/openhands... These ones are all built-in skills :) → tweet link
@sudoingX · 2026-05-26 05:59
let me say this out loud because nobody is framing it right. x premium+ is not a subscription anymore. it's the most valuable package in ai right now... grok build is a frontier coding agent on your CLI with real time x integration. → tweet link
@sudoingX · 2026-05-26 05:29
i woke up and went straight to grok build. this is now my peer agent. asked it to ssh into my nodes and check runs and updates. it did. reliably... xai is moving so fast it's hard to keep up. → tweet link
@gdb · 2026-05-26 16:48
magical experience with codex on iPad → tweet link
@gdb · 2026-05-26 15:51
Codex for analyzing and organizing your Slack: → tweet link
@gdb · 2026-05-25 22:09
Codex for finding space on your laptop: → tweet link
@jsuarez · 2026-05-26 18:03
Starting up the livestream, working on experiments for Puffer 5.0 and environment integrations. → tweet link
@nummanali · 2026-05-25 20:27
Talk about eating someone’s lunch. Google AI Studio lets you build native Android apps now → tweet link
@victormustar · 2026-05-26 13:29
As this blew up a bit here's the recipe to turn any model into a great Hugging Face Spaces demo autonomously 🚀 https://t.co/9RWxGqDBOo → tweet link
@TheAhmadOsman · 2026-05-26 13:11
The ultimate step-by-step projects roadmap for BECOMING an AI Researcher is now available online to read FOR FREE... You should read this, and if you cannot now then you most definitely wanna bookmark it for later → tweet link
@badlogicgames · 2026-05-26 08:24
RT @willmcgugan: Added support for @xAI 's Grok Build to Toad. Haven't used it in anger to build anything serious yet. But Grok Build look… → tweet link
Industry, Costs & Internal Politics
@kunchenguid · 2026-05-26 15:06
there is a mass misconception that $MSFT is cancelling Claude Code licenses because of the bill... it’s politics. having their own employees using claude code instead of copilot creates bad internal optics as well as external brand image → tweet link
@jezell · 2026-05-26 05:58
RT @CSGO500: 🚨THE AI COST CRISIS HAS STARTED. Valve reportedly told engineers to stop using Claude because AI bills were exploding, while t… → tweet link
@FinansowyUmysl · 2026-05-26 15:38
Świetna grafika pokazująca, dlaczego nie mamy do czynienia z bańką AI. https://t.co/q17G81E3m7 → tweet link
@badlogicgames · 2026-05-26 13:24
RT @edzitron: Uber’s COO has said that it’s getting “harder to justify” its AI costs because there was no way to show a link between AI spe… → tweet link
@TheAhmadOsman · 2026-05-26 18:26
Dario and Anthropic are just evil btw. Like, these are not things you don’t do without intentionally being evil and having an evil long term master plan → tweet link
@TheAhmadOsman · 2026-05-26 16:38
RT @MatthewBerman: "Anthropic’s public image is presently shaped by the idea that it “stood up to the Pentagon.” But the more nuanced versi… → tweet link
@Teknium · 2026-05-26 06:13
Our strategy lead @yeahfortommy was just on stage with CTO of @alibaba_cloud discussing Hermes Agent at the Qwen Conference, check it out: https://t.co/vU5G97OBUL → tweet link
@badlogicgames · 2026-05-26 08:25
RT @Jouhatsu_ai: Un ingénieur IA senior chez Microsoft vient de dévoiler comment les équipes de Microsoft créent des agents IA avec Anthrop… → tweet link
Software Development & Frameworks
@ASalvadorini · 2026-05-26 12:11
"I've used minimal and it's the real deal". Comments like this make me unbelievably happy... Minimal is the real deal 🔥, go check it out: https://t.co/NhT7TIsJxt #Flutter #Flutterdev → tweet link
@ASalvadorini · 2026-05-26 12:07
RT @techdocpcu: Google just broke their best AI tool. The Antigravity 2.0 update removed the IDE, killed the Cmd+E shortcut, and many more.… → tweet link
@steipete · 2026-05-26 11:13
RT @pwk: 𝗢𝗽𝗲𝗻𝗖𝗹𝗮𝘄 𝘃𝘀. 𝗖𝗹𝗮𝘂𝗱𝗲? 𝗧𝗵𝗲 𝗷𝘂𝗿𝘆 𝗶𝘀 𝗼𝘂𝘁!🏛️🥊 Last night, our 𝗚𝗿𝗮𝗻𝗱 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 𝗕𝗮𝘁𝘁𝗹𝗲 𝗠𝗲𝗲𝘁𝘂𝗽 issued the verdict with over 400… → tweet link
@badlogicgames · 2026-05-25 20:51
the one thing @mitsuhiko taught me: merged client & server logs. very useful. https://t.co/YOyxFV0zoa → tweet link
@ASalvadorini · 2026-05-26 17:18
RT @LucasJosefiak: Back from Google I/O. The future of @FlutterDev is bright ✨. Flutter remains #1 cross-platform framework and we could se… → tweet link
@badlogicgames · 2026-05-26 11:38
RT @GoogleDeepMind: SynthID has already watermarked over 100 billion pieces of content, but transparency is a team sport. That’s why we’re… → tweet link