Content About Stocktwits
No Image Available

Anthropic & OpenAI in Stop-and-Go US Traffic since ‘The Blip 2.0’ 3 weeks ago. ARD #109

Author
Michael Parekh AI: Reset to Zero
1 day ago
14 MIN READ

Today’s theme: Anthropic and OpenAI are stuck in stop-and-go traffic — almost three weeks since ‘The Blip 2.0’ — as the US government decides whether the latest frontier models are facing speed bumps, stop-and-go traffic jams, or brand-new speed limits. Three events I’d like to dig into for the AI Tech Wave — each with my Take first, then my Overall Take.

(1) Anthropic Finally Cleared to Re-Release Mythos & Fable 5 — After ~Three Weeks Shipping Other Software

MP TAKE: Anthropic reached a deal with the administration and should finally be re-releasing Fable 5 globally (with guardrails) starting this Wednesday — while Mythos stays gated. Released slowly to approved customers (first US organizations, then trusted corporate partners). This is almost three weeks since what I’ve called ‘The Blip 2.0’ three Fridays ago. (Named after ‘The Blip 1.0’ three-plus years back, when OpenAI’s non-profit board fired and then re-hired founder/CEO Sam Altman over a grueling weekend for the whole tech/AI industry.)

The tell of how long this dragged: while its frontier models sat gated, Anthropic spent the window shipping everything else — launching Claude Sonnet 5 (Opus-4.8-class performance at lower prices, “substantially better than Sonnet 4.6 for agentic work”), Claude Science (an AI workbench wiring Opus 4.8 into 60+ scientific databases and toolkits), and Claude Tag for Slack. Everything but the core models their advanced users worldwide were waiting for.

What both Anthropic and OpenAI need with alacrity is a clearer US-government policy on future major model releases — one with as little friction as possible, no speed bumps, no stop-and-go jams. To widen the lens: that’s exactly how Chinese AI companies are operating with their open-source models right now (event 3), and how US frontier labs themselves used to operate — right up until ‘The Blip 2.0.’ Remember, OpenAI’s ‘Blip 1.0’ took a weekend and a Monday. No such luck this time — and the longer the ambiguity runs, the more it costs.

Sources, in narrative order: WSJ — Anthropic reaches deal with Trump administration to restore access to Fable AI model. Anthropic — Redeploying Fable 5 · Department of Commerce clearance notice (X). Anthropic — Launching Claude Sonnet 5 · 9to5Mac — Claude Sonnet 5 cost comparisons. TechCrunch — Anthropic launches Claude Science (workflow over a new model). Anthropic — Introducing Claude Tag (for Slack). For longtime readers: ‘Anthropic flagged to a screeching halt globally by the US’ in AI-RTZ #1117.

(2) OpenAI ‘Previews’ a Next-Gen Model in the Mythos Class — But Isn’t Cleared to Ship Wide

MP TAKE: The next-gen preview from OpenAI — GPT-5.6, codename ‘Sol’ (which comes in large, medium and small, Sol being the big one) — is ‘nice to see.’ But since most of the market globally can’t actually touch any of it — it’s been available to only a handful of users — the benefits are theoretical. It’s not clear when OpenAI will be allowed to ship it wider; presumably soon, pending an announcement with the government.

This lands right as the industry is aggressively trying to move customers from free toward à la carte, metered pricing — but not GPT-5.6 at full blast, not yet. OpenAI is currently focused on streamlining its products on the AI-coding and ChatGPT fronts, fusing them toward an AI ‘Super-App’ aimed squarely at Anthropic — and it needs that fusion, because unlike Anthropic it has 900 million-plus weekly ChatGPT users worldwide, and it isn’t yet clear when 5.6 can power that business. To widen the lens — and to keep today’s metaphor running: it’s like the Formula cars are stuck in an LA traffic jam, engines revving, nowhere to go, the fastest machines on the grid throttled by the road, not their own limits. A preview you can’t broadly buy is a lap under caution.

Sources, in narrative order: OpenAI — Previewing GPT-5.6 Sol, a next-generation model. Axios — OpenAI releases powerful new GPT-5.6 model under restrictions. The Information — Trump administration asks OpenAI to stagger release of new model over security concerns. SemiAnalysis — TokenBudgeting: our conversations with enterprises on token spend. For longtime readers: ‘US AI Political Football Wars slow down US vs China’ in ARD #107.

(3) China Speeds Ahead Around the US AI Logjam

MP TAKE: China is accelerating its pace of open-source AI-model releases across its companies large and small — and claiming close parity with the latest US frontier models, including Anthropic’s Mythos/Fable — without the accompanying cybersecurity concerns that triggered the US slowdown. The standout is Zhipu’s Z.ai, whose open-source 5.2 model has gotten a lot of buzz precisely while everyone was twiddling their thumbs waiting on Anthropic and OpenAI; Alibaba and a raft of others are moving fast too. The FT frames it as China leapfrogging the US in the global market for open AI models.

Crucially, China is getting US customers in particular to actually try and compare the latest Chinese open-source models — which can be 50-80%+ cheaper and used in an un-metered way, without the à la carte pricing the US frontier labs want customers habituated to ahead of their mega-AI IPOs. To widen the lens: the US has far fewer open-source champions since Meta went more closed with its Muse Spark models (away from Llama). Nvidia and Apple, while best-positioned long-term to lead a US open-source AI answer, are not yet ramping as aggressively as they could — and the Chinese companies are taking full advantage of the US stop-and-go. The window for a US open-source champion is open now; it won’t stay open indefinitely.

Sources, in narrative order: CNBC — China’s Zhipu is closing in on top US AI models with Anthropic and OpenAI held back. NY Times — Chinese AI models close the gap with Anthropic & OpenAI. FT — China leapfrogs US in global market for open AI models. Axios — AI-access fight divides Trump’s tech allies. For longtime readers, in narrative order: ‘Anthropic’s Blip 2.0 in new form, as China AI ramps’ in AI-RTZ #1131; ‘US companies building on Chinese open-source LLM AI models’ in AI-RTZ #923; and ‘How Nvidia and Apple can be the global US open-source AI champions vs China’ in AI-RTZ #1089.

MP OVERALL TAKE

The stop-and-go approach of the US government around the latest frontier models from Anthropic and OpenAI could give their customers around the world reasons to try alternatives — especially on the open-source front. The key appeal for US customers of a ‘good enough’ open-source model would be un-metered and/or truly low-cost inference — even if it means spending aggressively on local, on-prem computing hardware — getting frontier-quality output without the ever-rising à la carte pricing OpenAI and Anthropic want to charge. That pressure is already showing up as ‘Token Budgeting’ among their customers large and small (SemiAnalysis), and open-source model use is increasing in the US, especially with Chinese models.

It’s good that the government has come to an arrangement with Anthropic to get Fable 5 out — that’s a terrific thing. But the pace at which the US unclogs the traffic jam, smooths the speed bumps, and gets the US frontier models running at high speed again in the coming days will determine how the AI Tech Wave does both fundamentally and financially in the markets ahead. Washington needs to clarify whether the US frontier models face speed bumps, interminable weeks-long stop-and-go traffic jams, or whether the speed limits are simply being dramatically reduced for the best US models. Those are critical questions for investors and — more importantly — for customers weighing US offerings against the opportunities from companies around the world. In the worst case, China gets to race ahead of the US globally in this AI race — while US companies rev their engines in a traffic jam, waiting for clarity on the new rules of the road, and the new speed limits.

Gadget AI — Quantum Computing: ‘Good For Absolutely Nothing… Yet’

MP Take: I want to talk quantum computing — and the reason it belongs in an AI context is that Nvidia and others are now using AI to accelerate quantum research. The Verge had a detailed piece whose title says it as pithily as possible: “What is a quantum computer good for? Absolutely nothing — yet.” Despite all the genuine technical innovation still underway — and as fascinating as it is at a deeply technical level, and as profitable as it may be in the markets while investors race way ahead of the realities in a bull market with Microsoft, Google, Amazon, IBM and others pouring billions into bets years in the making — the honest status today is closer to “good for absolutely nothing… yet.” These technologies are at a very nascent stage; the markets are compressing what’s likely a 5-10-15-year arc into 2-3-year expectations.

To widen the lens, the irony is almost too neat: quantum computing, when it finally arrives, was long supposed to be the ultimate cybersecurity nightmare — the RSA-encryption cracker that would theoretically undo the world’s computer-security, financial and corporate systems. Instead, the thing that actually broke the AI industry’s stride on cybersecurity this month wasn’t quantum at all — it was Anthropic’s Mythos/Fable and OpenAI’s GPT-5.6, both sitting stalled and revving in traffic on the AI road. What we were waiting for on one side happened on the other. The feared future threat is theoretical; the real one showed up as a policy speed bump.

Sources, in narrative order: The Verge — What is a quantum computer good for? Absolutely nothing — yet. For longtime readers: ‘Microsoft achieves a milestone in Quantum Computing’ in AI-RTZ #638.

Questions

Q1 — What does MP find truly amazing about quantum computing’s technical realities so far?

I’m a Physics and Quantum Mechanics nerd — a lifelong love — so the deep science here is utterly fascinating. It genuinely dances with the still-unknown realities of quantum-mechanical physics, and there’s an amazing amount still to be learned in both the basic and applied sciences around this technology. That’s what draws me to the field — even as we remain far away from the application side.

Q2 — What’s the sobering reality, from MP’s perspective?

As amazing as the basic science keeps being, the applications will likely take longer than the financial markets assume — the familiar pattern of a bull market bidding up the stocks (no financial advice here), opening a widening gap between the science and the expectations. The key quote from the Verge piece says it: “So far, this era of quantum-computing science is complete codswallop,” — Henry Legg, a physicist at the University of St Andrews. I rather like the phrase.

Full Source Reading —

For the broader context, see the canonical sources for ARD 109 — in today’s narrative order:

Event 1 — Anthropic Cleared to Re-Release Mythos & Fable 5

Event 2 — OpenAI Previews GPT-5.6 ‘Sol’, Not Cleared to Ship Wide

Event 3 — China Speeds Ahead Around the US AI Logjam

Gadget AI — Quantum Computing

Shorts Clips from today

Clip 1 — US Needs Clear AI Model Approach

Watch on YouTube Shorts

Almost three weeks into ‘The Blip 2.0,’ Anthropic and OpenAI are in stop-and-go traffic while the US government decides how their latest frontier models can ship. What both labs need, with alacrity, is a clear release policy.

MP Take: Least friction possible — no speed bumps, no weeks-long jams — the way Chinese open-source labs operate now, and the way US labs used to. The pace at which Washington unclogs the jam determines how the AI Tech Wave does, fundamentally and financially. In the worst case, China races ahead globally while US companies rev their engines in traffic.

Clip 2 — AI’s Surprising Cybersecurity Threat

Watch on YouTube Shorts

Quantum computing was long supposed to be the ultimate cybersecurity nightmare — the RSA-encryption cracker that could undo the world’s security, financial and corporate systems.

MP Take: The irony is almost too neat. Instead, the thing that actually broke the AI industry’s stride on cybersecurity this month wasn’t quantum at all — it was the frontier models themselves, Anthropic’s Mythos/Fable and OpenAI’s GPT-5.6, stalled and revving in traffic under US-government gating. The feared future threat is still theoretical; the real one showed up this month as a policy speed bump.

Clip 3 — Quantum Computing: Good for Nothing Yet

Watch on YouTube Shorts

The Verge’s pithy title says it all: “What is a quantum computer good for? Absolutely nothing — yet.” And Nvidia and others are now using AI to accelerate quantum research.

MP Take: Genuine innovation, fascinating at a technical level, and profitable in a bull market as Microsoft, Google, Amazon and IBM pour billions into bets years in the making — but the honest status is ‘nothing yet,’ with the markets compressing what’s likely a 5-10 year arc into 2-3 year expectations. The key quote: “This is complete codswallop,” per physicist Henry Legg (University of St Andrews).

About AI Ramblings Daily (ARD), and AI-RTZ

Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.

AI-RTZ is the morning text — a deeper written take on one idea, published by at least 5 AM EST. Today: post #1134.

AI Ramblings Daily is the afternoon video + podcast — my ad hoc takes and perspective on the day’s AI issues & news flow, around 20 minutes, with short 1-2 minute clips for quick topic views. Today: episode #109.

Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.

Links used in today’s show (already embedded inline above; listed here for reference)

Take 1 — Anthropic Cleared to Re-Release Mythos & Fable 5:

Take 2 — OpenAI Previews GPT-5.6 ‘Sol’:

Take 3 — China Speeds Ahead Around the US AI Logjam:

Gadget AI — Quantum Computing:

Q1 + Q2 — MP on quantum computing:

  • (no external sources beyond the Verge piece — MP’s own analyst view)

Companion text:

AI Ramblings Daily on AI-RTZ is here to think through AI and reset. Together.

Today’s AI-RTZ #1134 — Latest report on an AI Bust & the Global Economy — on a new Bank of International Settlements report modeling what an AI-bubble burst could mean for the global economy, with some interesting details and charts. Recommended as today’s reading post.

Tomorrow — ARD 110 on AI-RTZ 1135.

Thanks for joining us today, AI Curious Folk. Stay tuned.

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)

Subscribe to AI: Reset to Zero for daily AI Ramblings + Sunday Bigger Picture posts.





Want the latest?

Sign up for Michael Parekh's Newsletter below:


Subscribe Here

More like this

BROWSE ALL ARTICLES