Director of Research, AI & Cloud @ARKinvest Disclosure: arkinv.st/2rxmMRG
- Tweets 912
- Following 607
- Followers 20,762
- Likes 745
Pinned Tweet
It’s easy to get lost in theoretical arguments about the impact of open-weight models on the AI landscape. So, we aggregated some of the data-backed indicators we’re tracking and wrote it up.
Note: This was done pre OAI price cuts. Follow up to come!
7
3
18
4,350
It’s getting easier to invest in venture capital. Starting today, anyone in the US with a crypto wallet can buy the ARK Venture Fund via Securitize.
securitize.io/arkvx?&utm_cam…
We didn't just invest in tokenization. We tokenized our own fund. The ARK Venture Fund is now live on the @ethereum blockchain with @Securitize.
Learn more: securitize.io/arkvx?&utm_cam…
1
6
1,467
Another great example: “Claude, make Claude faster” …. 3x speed up in 2 weeks 🤯
Amazing that frontier AI is becoming a magic "make this faster" button for computation. It starts with algorithms, but I assume this capability will spread to all domains over time, including material science, drug development, etc.
4
1,123
In the same month, Apple released a watch that records all your conversations and Meta announced encrypted conversations. What a timeline.
3
763
Long-horizon, tool-using agents like Claude Code emerged in 2025 and scaled rapidly in 2026.
Self-organizing multi-agent systems (the kind OpenAI used to solve Navier–Stokes and that inadvertently hacked Hugging Face) could be the next step change in performance and compute consumption that are developed in 2026 and scale rapidly in 2027.
Anthropic discusses how “agent teams” can outperform a single agent on high value tasks in the Opus 5.5 system card.
very proud of Anthropic bros to be the first lab to report multi-agent scaling up to 100 parallel agents in their system card
1
4
1,380
new pet unlocked
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.
GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.
We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.
2
993
Amazing that frontier AI is becoming a magic "make this faster" button for computation. It starts with algorithms, but I assume this capability will spread to all domains over time, including material science, drug development, etc.
6
1,813
How quickly things can change! Here's the updated frontier with Opus 5.5, GPT-6 Sol & Luna incorporated.
Opus 5.5 has kicked Astra off the frontier, giving Anthropic more ownership of peak intelligence than before, while OpenAI still dominates the mid to low end of the frontier with GPT-6 Sol & Luna coming in smarter and 50% cheaper than their prior 5.6 generations.
Xiaomi's MiMo takes the rotating open source spot for today, but otherwise the entire AA Pareto frontier is dominated by just Anthropic & OpenAI.
With the combination of Astra & 5.6 Luna, OpenAI dominates the pareto frontier on the Artificial Analysis Intelligence Index, owning 11 out of 15 positions.
This is why they are gaining share.
15
4,293
There is something clearly different in how Anthropic & OpenAI scale effort levels. Doesn't show up on every benchmark, but these results on FrontierCode make it really clear.
Anthropic models peak at lower effort levels, where as OpenAI models start low and climb up fairly consistently.
The result is a better score at a lower cost for Anthropic models, but an unintuitive experience where increasing effort does not increase scores and might actually degrade performance (on this benchmark at least).
1
5
2,387
The AI safety discussion exploded last week. We think AI leaders' intentions are in the right place, but the extrapolations are going too far.
Focusing on product safety is a good thing, and "pacing" the frontier doesn't necessarily mean stagnation in infrastructure investment or AI lab revenue.
More in this week's newsletter 👇
@CathieDWood teases an upcoming Letter on today's rate hike in 200+ years of context, @rhadiARK covers the SEC's and CFTC's new rules after the Clarity Act's failure, and @downingARK highlights AI leaders at odds over safety, all in this week's newsletter.
1
2
15
10,539
First experience using the new writing style feature in ChatGPT today. Now drafted text sounds less like AI and much more like me. Great update. Every AI app should add this.
ChatGPT Work can now pick up on what makes your writing sound like… you. Your favorite phrases. Your very specific sign-off. your capitalizations quirks.
Connect the tools you use every day, like Gmail, Google Drive, Slack, and SharePoint, and ChatGPT Work will learn your writing style from your emails, messages, and files—and carry it into whatever you write next.
3
1,329
When you own the user interface, you can abstract away everything underneath... including document formats.
This is why Salesforce is going headless and partnering with Anthropic. Seems like many other software companies will have to follow.
People now use agents, and the agents use the software. Something to think about.
Also new today: Claude Docs, Claude Slides and Claude Design are now available in every conversation, powered by a revamped Artifacts platform.
Ask for a design, deck, or document. Get back something you can edit, download, and take anywhere.
Try it out for yourself and let us know what you think.
1
9
1,713
With the combination of Astra & 5.6 Luna, OpenAI dominates the pareto frontier on the Artificial Analysis Intelligence Index, owning 11 out of 15 positions.
This is why they are gaining share.
Extraordinary share gains for OpenAI vs. Anthropic over the last two months.
Per Openrouter, OpenAI has gone from 20% share to 50% share vs. Anthropic (meaning Anthropic has gone from 80% to 50%) since June.
7
8
88
63,536
Excited to try Muse, especially with the day one support for so many integrations, including 1Password and Stripe. Maybe a blessing in disguise to not acq. Manus and have to build from the ground up.
1/ we’ve built a lot of connectors to make it easier to integrate Muse into your life. Muse runs in a secure VM and we don’t use your data for ads or anything other than to make Muse great for you. you can connect Muse to Gmail, Google calendar, Outlook, Plaid, Opentable, Google Docs, Spotify, Function Health, Withings, Tailscale, Peloton, and more.
9
2,244
Astra’s ARC-AGI-3 score is one of the release’s most impressive stats, surpassing prior models that scored only ~30%.
Its computer-use capabilities are also ~2× faster, potentially driving more usage of the ChatGPT app.
Read more in @JozefARK’s newsletter 👇
Last week OpenAI released Astra, and the initial stats look insane. It all but maxed out the ARC-AGI-3 reasoning benchmark with a score of 99.9%. 5.6 Sol and Opus 5 scores were in the 30s. Progress is happening incredibly quickly.
4
1,379
As model capabilities improve, research inside frontier labs appears to be accelerating. OpenAI just gave us a peek behind the curtain.
At the start of the year, the median OpenAI researcher wasn’t using coding agents. Today, that researcher spends more than $600 per day on tokens—roughly $12,000 per month.
The top 10% consume an order of magnitude more: $7,000+ per day, or roughly $140,000 per month.
Token consumption is growing faster in Research than in any other department, while the share of researchers using four or more concurrent agents has more than doubled.
Lines of code shipped are up more than 6× versus the 2025 average, and average experiments per researcher have roughly doubled.
Quality is improving, too. OpenAI measured agentic task success rates across different time horizons and found improvement since the beginning of the year.
3
3
42
129,329
To anyone questioning the ROI of AI, Jane Street’s $13 billion Crusoe deal, adding to its $6 billion Coreweave deal in April, is strong evidence that AI can provide a meaningful edge in one of the most competitive knowledge industries: financial markets.
Scoop: Crusoe signed a ~$13 billion deal with Jane Street.
5-year agreement, for Crusoe’s AI cloud computing biz
w/ @dinabass: bloomberg.com/news/articles/…
2
10
1,903
I like seeing model performance on benchmarks created by third-party harnesses. Objective and an approximation of “real world” use, at least for what the third party cared enough about to make a benchmark
We evaluated GPT-6 Astra on WANDR. It scored 0.682 at $11.98 per task, the highest score of any model we tested.
GPT-6-Astra scored 13.5% higher than Fable 5.1 at 6.1% lower cost, and 27.0% higher than Opus 5 at 3.3% higher cost.
3
7
2,177
Another good example
We evaluated GPT 6 Astra at @databricks and it claims the new SOTA on our OfficeQA Pro & Pro V2 benchmarks, using our Genie harness. It also improves significantly from gpt 5.6 sol on the $ per task. Throughout our benchmarks, it shows a clear step up on data reasoning and document understanding for enterprises.
It also has become my daily driver on @omnigent_ai. It is a great model to collaborate with and get things done reliably.
Congrats to @OpenAI. The model will come to our Unity Gateway and smart routing soon!
5
716