@akramsrazori
iAccount based inWest Asia
About this account
- Account based in
- West Asia
- Connected via
- North America App Store
Account-level information from X, not a live location or the device used for a specific post.
Hyyypergrowth, SaaS, Digital money enthusiast, CEFI Skeptic, Oil bull, and Futurist. If you ain't long, you're wrong. Occasional Security Analysis.
United States
Joined December 2011
- Tweets28.4K
- Following740
- Followers24.8K
- Likes16.7K
Pinned Tweet
If tobi had just been following me on twitter this $shop drama all could have been avoided.
the-razors-edge.ghost.io/raz…
Jev showing that some decent fraction of what we r currently calling llm inference was never fundamentally a language gen workload at all. U cud have said was always obvious and that at mature workload production llms wud written lots of deterministic soft already and handed off, but this crystallizes that at ai speed. Basically we moving past first gen agent architecture fast.
Replying to @sunglassesface
Top 10 Jev use cases in agentic apps, ranked by theoretical gain (speed/cost/reliability) over LLMs:
1. Real-time loops (games/robotics): enables 10+ decisions/sec; LLMs too slow.
2. Browser/computer-use action selection: 100x+ cheaper/faster multi-step runs.
3. Tool risk gating: instant safe/unsafe per call; zero hallucination.
4. Model routing: routes cheap vs frontier; massive cost cut.
5. Goal/stuck checks: sub-second loop control.
6. Context compaction: keep/delete decisions beat lossy summaries.
7. Skill/tool selection: precise activation without prompt bloat.
8. Output/trace guardrails: calibrated jailbreak/policy checks.
9. Ticket/email triage: high-volume routing at 400x lower cost.
10. RAG reranking: relevance scores over candidates.
Jev returns typed probs in 70-500ms; LLMs generate text slowly/expensively for same decisions.
DaRazor retweeted
Bo Jackson won the state decathlon title in high school doing only 8 events!
Was afraid of the pole vault and hated the mile run, so just got enough points in the other events to beat everyone else.
No one comes close to him as an athlete. Imagine Derek Henry or Aaron Donald playing MLB and hitting home runs and throwing runners out at home from Right Center.
DaRazor retweeted
Jev dropped the price of SEO/GEO fixes by 90%
Agents that audit and fix a client's SEO/GEO used to cost us ~$250
Here's where the savings come from:
1/ 30x faster reads of Search Console and PostHog/Mixpanel data
2/ 30x faster checks of what ChatGPT searches on Bing
3/ 30x faster modeling of what users ask Gemini and Claude
4/ 30x faster scans of who ChatGPT and Claude cite
5/ 30x faster analysis of the sources behind those citations
6/ 30x faster gap analysis: why they get cited and we don't
7/ 30x faster fixes across 1,000s of pages on large client sites
8/ 30x faster sorting of which page types ChatGPT cites
9/ 20x faster creation of the pages that make ChatGPT pick you
Available in the Ryze AI app and MCP/Claude Connector, link in the 1st comment 👇
Jev u son of gun u
Jev is WILD for SEO audit 🤯
in 45.1 seconds it read all 586 pages on my site and rebuilt the internal link map. 584 links placed, 139 pages it refused to link because nothing honestly fit. total cost $0.21.
Claude Opus 5, same 586 pages, same clock, got through 21 of them and spent $1.43.
per page that is ~190x cheaper. the full Opus pass would have run $43.
internal linking is the perfect Jev job. it is not writing, it is 8,790 yes/no calls: does this page have a real reason to link to that one, and is there anchor text already sitting in the copy. that is a classification problem, and we have been paying frontier prices to do it one page at a time.
what you are watching: left column is Jev, right is Opus, same queue, same rubric. the run stops the moment Jev finishes so Opus stops burning tokens.
coming soon to @distribb_io + connector + gpt plugin
“It’s the Americans; they want to organize a counterattack on the llms. They claim to be able to collapse their inference costs.”
youtu.be/xqRGx4y4J24?si=_UZ4…
Optimize Prime
Jev was adopted faster than any other model in AI Gateway history.
In the first day, @typesafeai reached ~13% of teams, 2x the GPT-5.6 family and 6x Fable 5.1.
Deflate w prejudice
1/ Introducing CUA-S1: a family of System One Models, small, specialized, and built for computer use.
Today we're open-sourcing CUA-S1-FORMS, the first in the family: github.com/trycua/cua
All capex life portfolio jizzing aside think we might need to understand reward hacking and rl system objective optimization just a little bit better. Like if u can attach lethal force to an agent in a robocop or autonomous drone and u tell the agent kill x and minimize civilian causalities, its a serious problem if it optimizes that 500 casualties is ok. And better question do we really believe this isn’t already out there.
Opening access for developers to build Muse connectors. You bring the API -- Muse brings the agent, the browser, and the context of what the person actually wants. People reach your service just by asking for it, and their agent takes it from there.
New connectors are live today. Come build with us. muse.ai/platform
DaRazor retweeted
Drama ads aren’t just for women.
That might be one of the dumbest assumptions in advertising right now.
Men will watch drama.
You just have to give them a reason to care.
And one of the easiest ways to get a man emotionally invested is to put him in a situation that hits his ego.
He feels himself getting weaker.
His hair is disappearing.
He can’t keep up with the younger guys anymore.
His wife doesn’t look at him the same way.
He’s embarrassed by something he would never admit out loud.
That’s your hook.
But the mistake is revealing the product immediately.
Let the story breathe.
Let the insecurity build.
Let the viewer wonder what happens next.
Then, at the end, reveal the bigger picture.
Because the story was never really about the product.
It was about confidence.
Identity.
His relationship.
Getting older.
Feeling like himself again.
That’s what makes drama so powerful.
Almost any product can become a prop inside a story that actually means something.
The product is the mechanism.
The story is the reason to watch.
And this is where watch time becomes a massive creative advantage.
If you can get someone to watch 30, 45, 60+ seconds instead of 2 seconds, you’re creating dramatically more opportunity for Meta to learn who is actually responding to your creative.
More watch time.
More engagement.
More shares.
More signals.
And potentially more efficient acquisition as the system finds people who respond to that type of content.
That’s the part people miss.
The goal isn’t just to make an ad people click.
The goal is to make an ad people watch, finish, and send to someone else.
Because when the story earns the attention, you get something traditional UGC rarely gives you:
Time.
And time is one of the most valuable things you can get in advertising.
You just can’t directly buy it…
People can just swipe right past your ad…
You have to convince the customer to give it to you.
This is very interesting. Feel like these decision models will eventually be big deal in agent flow or anything that is high vol repitive call driven.
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
$META is starting to put a subscription layer across its entire ecosystem with Meta One now live globally across Instagram, WhatsApp and Facebook.
Plans start at $7.99 for consumers and reach $499 for businesses with more AI usage and premium tools bundled in.
$now this really shud be a lot higher already...the relative gap remains too wide...like fine dont give it a cyber or grc bid but its gotten stupid beyond that as well
Will $meta be the first quadrillion company or the last trillion company?