Tracking the AI frontier: models, agents, research, leaks & pre-release signals. I test, investigate and build.

Tips, leaks & collabs → DM
Joined March 2022
Gave two Flash models the same prompt: build a storm at sea with a lighthouse in Three.js. DeepSeek V4.1 Flash went for a lighthouse. MiniMax M3.1 Flash went for a movie scene: a sweeping beam through the fog, lightning inside the clouds, foam breaking on the crests. Then there's the bill. Both start at the same price: $0.15 per million input tokens and $0.60 per million output. But DeepSeek doubles to $0.30 and $1.20 at peak hours on weekdays. MiniMax M3.1 Flash stays flat, any hour, any day. That sounds small until you run agents. For a single prompt, nobody cares. For agents that run for hours, it's the difference between leaving it on overnight or babysitting the clock. With flat pricing you can iterate, retry and leave it running without checking the clock. Quick tip for existing M Plan subscribers: you can use MiniMax Code without limits until October 25. I'd use that window to stress test M3.1 Flash on a real project before deciding anything. Links below.
4
14
1,809
Links if you want to test it yourself: agent.minimax.io/ minimax.io/m-plan
2
475
Grok Bot runs on Opus 5.5, and Claude Duet runs on Opus 5.5. So we literally have two competing products powered by the EXACT SAME MODEL... It's Anthropic competing against Anthropic itself, using SpaceXSI's compute, while also taking on OpenAI. This industry is absolutely insane.
Replying to @claudeai
We designed Claude Duet around Claude Opus 5.5, which gives it the judgment to decide what needs you, what can wait, and when to check with you first. It carries over the memory, skills, and connected apps from your Claude account, so there’s little to set up.
30
9
2
350
17,988
"I smell fear..." Tibo definitely felt that Claude Duet launch. Wanna fight back, Tibo? Bring Dots to Europe and give Plus users access already!
Dot, Dotter, Dottest
25
7
299
10,881
THIS WEEK HAS BEEN ROUGH, GUYS. EVERYONE'S GETTING HACKED. ☠️😭
Railway is compromised, rotate all API keys as soon as possible Deployments are down so don’t wait to redeploy Attackers are draining keys in real time, we didn’t lose much but can confirm this is still an active attack Stay safe friends
4
98
5,682
WAIT, WHAT?! Anthropic just decided to OBLITERATE OpenAI Dots and Grok Bot, dropping Claude Duet powered by the absolute monster that is Opus 5.5. We literally go to sleep, wake up, and Anthropic has launched ANOTHER product to crush its competitors. THESE GUYS JUST CAN'T STOP WINNING. WHAT AN ABSOLUTELY LEGENDARY RUN.
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent. It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
161
76
13
2,657
489,734
Token Gremlin retweeted
Gemini 4 carbon test check point 1: ‘gemini 4 argon’ in arena
does anyone still use Gemini 4 argon ? if yes, google is preparing to launch Gemini 4 carbon early access partners have started receiving Gemini 4 Carbon EAP and they’re saying there’s a noticeable improvement from Gemini 4 argon cons: it is slower and bad at frontend testing now in arena wish me luck 😂
18
4
2
89
8,379
I thought it was Fable 5.5, but nope. Just Anthropic telling us we can now use Claude to update our code in real time to defend against attacks from other Claude models that people will inevitably use against us. (AND TRUST ME, THEY WILL!)
Claude Security now keeps up with your code as it changes. Connect your repositories once. Claude runs a deep first scan, then re-scans as changes land, so findings stay close to the current state of your code. In public beta on Claude Enterprise.
11
4
324
22,904
So basically, Fable 5.5 is already the best model in the world at pretty much everything, and now it's going to solve experimental physics problems too. Anthropic just can't stop winning.
Dr. Christopher Stubbs is joining our Long-Term Benefit Trust. Chris is an experimental physicist and a former Dean of Science at Harvard University. He has also advised the US government on national security and arms control. Read more: anthropic.com/news/christoph…
17
20
1
683
26,123
Almost 70% of the web compromised, humans and AI working together to fix it, and literally nobody in my everyday life has a clue what I'm talking about.
41
115
8
1,740
25,857
Token Gremlin retweeted
Two critical vulnerabilities have been identified in PHP, a widely used programming language that powers websites and web applications. These vulnerabilities (CVE-2026-109618 and CVE-2026-109619) could allow malicious cyber actors to remotely execute code in PHP core, potentially compromising sensitive data and networks. The FBI urges organizations to take the following steps now: ➡️ If an immediate update is not feasible it is recommended that organizations take the system offline until it can be fully mitigated. ➡️ Immediately update affected PHP installations following the guidance on php.net. Inventory where PHP runs in your environment, prioritizing internet-facing systems and vendor products that bundle PHP. ➡️ Monitor web servers for suspicious activity, including PHP processes unexpectedly launching command shells or downloading files. Ensure appropriate process-monitoring capabilities are in place to help identify potential exploitation. ➡️ If you find signs of compromise, isolate affected systems, preserve your web server and PHP logs, and report it to your local FBI field office or ic3.gov. Those reports help us identify the actors involved and warn other organizations before they become victims. For the latest information, follow PHP’s announcement on php.net, CERT-EU Security Advisory 2026-018, and CISA’s Known Exploited Vulnerabilities catalog.
343
4,537
989
21,822
2,358,732
Token Gremlin retweeted
The PHP security team is giving advance notice: Upcoming security release Wed, Oct 21st 2026, 14:00UTC, two critical vulnerabilities, both rated CVSSv4 9.2. PHP users are encouraged to prepare to update. Earlier release if details become public or active exploit evidence.
335
4,125
1,234
18,240
6,809,001
You wake up and suddenly find out that almost the entire web is exposed. What a time to be alive.
5
7
197
6,335
Token Gremlin retweeted
The PHP project has released fixes for 2 critical vulnerabilities affecting servers that host PHP applications. Site owners and technical teams should confirm their hosting provider’s update plan and update PHP immediately. Versions before PHP 8.2 no longer receive security updates. These issues are not specific to WordPress, and no WordPress update is required. wp.me/pZhYe-5Gb
79
1,207
136
3,274
275,370
Another one from OpenAI, and this is exactly the future I keep talking about: Thousands of GPT-Rosalind agents analyzed data from millions of mouse brain cells in 32 minutes. The team estimates the same work would take humans months to a year. They surfaced new leads for Parkinson-like disease and tuberous sclerosis, and experts called them novel and worth testing. The best part: thousands of agents in parallel cited 700+ papers. One agent alone cited 11. PHP this morning, the brain this afternoon. Scale the agents and we'll be finding things in months that would've taken us decades.
GPT-Rosalind surfaced a new lead for drug discovery: could restoring cholesterol transport protect neurons from degeneration? DCTN1 is a gene linked to Perry syndrome, a rare neurodegenerative disorder. In mouse-brain data, disrupting it increased expression of genes involved in making cholesterol. GPT-Rosalind suggested a possible explanation: cholesterol may not be reaching where neurons need it. That gives scientists a specific mechanism to test and, if confirmed, a direction for developing new treatments for neurodegenerative disorders. With @perturbai_tx and @scrippsresearch, a team of thousands of GPT-Rosalind agents worked in parallel, connecting gene-expression results from millions of mouse brain cells with scientific literature in just 32 minutes. We’re now expanding this collaboration to identify new drug targets. openai.com/index/scaling-ai-…
6
9
2
166
8,745
"An extremely urgent request to everyone. We ask that you take action tonight. The PHP Security Team has issued an alert regarding two critical vulnerabilities in PHP, which is widely used across websites. Please check whether your public-facing websites use PHP and prepare to apply the security patches as soon as possible after their release tonight, October 21, at 11:00 PM Japan Standard Time (JST). If you are unable to apply the patches promptly, temporarily taking your websites offline may also be an effective measure to reduce the risk. Japan's National Cybersecurity Office is also sharing relevant information through its X account. Please refer to their updates as well. National Cybersecurity Office (Alerts and Advisories): @cyber_forecast"
皆様に、今夜に対応して頂きたい大至急のお願いです。   多くのウェブサイトで使用されているPHPに関し、PHPのセキュリティチームが、重大な2件の脆弱性に関する注意喚起を行っています。   公開しているウェブサイトでPHPを使用しているか確認し、今夜、10月21日午後11時(日本時間)に公開される修正パッチについて、可及的速やかに適用できるよう準備をお願いします。   修正パッチを速やかに適用することが困難な場合は、例えばウェブサイト等を一時的に停止させることも、リスクを低減させるための有効策の一つと考えられます。   国家サイバー統括室のXアカウントでも関係する情報を発信していますので、そちらも参考にして下さい。 国家サイバー統括室(注意・警戒情報)(@cyber_forecast) / X
6
11
123
8,121
Good morning, friends. Incredible news from OpenAI: Roughly 70% of the web runs on PHP, and researchers just used GPT-5.6 Sol to find a critical security flaw in it (yes, 5.6, we're not even talking about Astra or Bel yet!). And OpenAI says this is just one of several serious vulnerabilities it has already found. Makes me wonder how many more we'd uncover with a few hundred thousand AI agents reviewing all the code humans have ever written.
OpenAI models helped researchers—operating as part of our Patch the Planet initiative—discover critical security flaws in PHP, an open-source programming language used to run an estimated 70% of websites. Over the past two months, we led a coalition of maintainers, researchers, and security providers to prepare fixes and protections ahead of disclosure. This duration was necessary because the fix was quite intricate and required a major change to the PHP engine. Every site using PHP can now safely upgrade their version to stay secure, in addition to applying mitigations from various security vendors. Due to the severity of these flaws and PHP’s widespread use, we consider this bug to be one of the most impactful vulnerabilities in recent memory. It also marks a new phase, where AI can find security bugs with internet-scale impact. This is an example of the Defender’s Window in action: to our knowledge, frontier AI was able to find this vulnerability ahead of attackers, giving defenders time to prepare. openai.com/index/coordinated…
Readers added context they thought people might want to know
Wiz Research wasn't the first one to discover the exploit. An independent researcher first discovered it: x.com/calif_io/statu…
9
7
235
16,474
Claude Code with Sonnet 5.5 on Max managed to outperform Devin Fusion running Opus 5.5 Max and GPT-6 Luna High together! And that's saying A LOT, because Devin Fusion is genuinely amazing to use. So I can only imagine how incredible Sonnet 5.5 on Max must be in Claude Code right now, especially with those almost unlimited usage limits depending on your plan.
Devin Fusion with Claude Opus 5.5 (max) and a GPT-6 Luna (high) sidekick scores 68 on the Coding Agent Index, 2 points above Claude Code with Opus 5.5 (max) at 31% lower cost per task @cognition's Devin Fusion runs a frontier lead model alongside a lower-cost sidekick model. The two work in parallel with their own context and tools, with the lead keeping planning, delegation and final review. We benchmarked Devin Fusion CLI with Claude Opus 5.5 (max) + GPT-6 Luna (high). It scores 68.1 on the Artificial Analysis Coding Agent Index, closely matched at the top with Claude Code with Claude Sonnet 5.5 (max) at 68.4 and Muse Code with Muse Wonder Full (max) at 68.0. Against Claude Code with Claude Opus 5.5 (max) at 66.0, it costs $9.01 vs. $13.04 per task and averages 55 vs. 65 minutes per task. Across the three benchmarks, the Fusion configuration scores higher on DeepSWE v1.1 (72.6% vs. 68.4%) and SWE-Atlas-QnA (69.6% vs. 66.4%) and slightly lower on Terminal-Bench 4.0 (62.1% vs. 63.1%). Its results reflect both models working together, including their combined API cost. Costs are pay-per-token API costs, not subscription prices.
19
2
91
10,494
People love Google way more than I expected. But I'm not sure how they measured this. If we're talking specifically about Gemini 4 Argon usage, then it was a MASSIVE success.
In just the first 72 hours since Gemini 4 Argon launched, we processed ~25 trillion tokens, the demand for intelligence is incredible and still under appreciated!
11
4
1
154
8,509
Looks like Google's real challenger to Opus 5.5, Gemini 4 Carbon, is getting close to launch. I'd say we could see it in the next few weeks. And honestly, I'm pretty excited. Gemini 4 Argon is already a really good model, so I'm curious to see how much further Google can push things with Carbon.
Gemini 4 Carbon EAP has officially started. Early access has been given to partners and it's being tested in LMArena. The release is very close. It's noticeably better than Gemini 4 Argon at 3D, though frontend/UI is still bad. It's also slower but less lazy.
10
3
1
215
10,338