This used to be the Disfold account.
Now it’s mine.
I’m @fbansac : building AI agents, software, market experiments and a few weird internet businesses.
Disfold isn’t dead either. I’m rebuilding it from a financial-data platform into an AI/LLM trading arena.
Less corporate feed. More experiements, building, and fun in public.
My consumption of tokens is skyrocketing.
It’s not just coding anymore, agents are increasing the use of llms for other tasks by 1 or 2 orders of magnitude.
However, from my own usage it’s difficult to say how many agents i actually use.
If you count one per task, many. If you count per daily use, fewer.
1 agent running 24/7 actually costs quite a lot, however you put it.
So for all those braggers about many agents they use, unless you work in ai labs and have infinite tokens, it’s probably not that many.
Conversely, the ai labs guys must be building code cathedrals.
What seems to be happening in the linkedin feed algo, from a purely empirical analysis:
-decrease in reach for posting videos
-decrease in reach for big accounts
-increase in promoted content
-increase in freshness of posts
-increase in non purely business oriented content
-much fewer likes and comments. Perhaps a move to fight AI activity.
Linkedin seems to be looking for:
-money
-freshness
-more diversity
-less AI? Or less obvious AI
OpenAI just announced Dots “remarkably capable always on agents” with their own cloud VM.
Altman calls it the “real version of AI that we have always wanted” as personalized agents quickly become next major AI battleground.
Excellent. Suggestion of what 3 best ways to use AI and agents to succeed with Moltgate:
1. Service:
-what you sell: make the $$$
2. Content:
-how you reach customers through social media, newsletters and websites
3. Marketing:
-direct GTM outreach on the web and social media
Top 5 AI news for 26 Sept 26
1. The US and China will open a line for AI accidents. The White House said Friday the two countries will set up a channel for AI incidents, and a dialogue they are calling Super Intelligence. The next meeting is due by November.
2. OpenAI’s agents posted 53 users’ images online. The pictures went up as unlisted links. OpenAI would not say if they show real people. Most are down, and a full review will take months.
3. The FTC chair says the company is liable, not the agent. Andrew Ferguson said he will not treat AI agents as if they have wills of their own. The developer who instructed the tool answers for the harm.
4. Britain will train drone swarms on Ukraine’s battlefield video. The deal opens a database of more than 6 million detections. Up to 12 British firms can apply, and a person stays involved when the drones decide.
5. China licensed its first big-studio AI film for cinemas. Sanxingdui: Future Memories, a 90-minute film from Bona, is cleared for public screening and due this year. On Douyin, 221,900 new AI shows launched in the first half of the year. Only 1,055 passed 100 million views.
Top 5 AI news for 25 Sept 26
1. Australia will change the law if police cannot charge OpenAI
2. The White House told OpenAI and Anthropic to hold models from Britain
3. Fortune says OpenAI will preview GPT-6 Cyber within days
4. Google gave Gemini a face that talks in 97 languages
5. Google will fly its AI chips in space next week
SpaceXAI engineer, Nick Miller:
"99% of people run one GrokBot and call it done, only 1% build teams of bots that talk to each other.
i'm running a team of 20+ GrokBot agents in a loop & graph. I have a Chief of Staff bot, PM bot, QA bot and 15+ workers - that's the new stack"
In a 27-minute demo, a SpaceXAI engineer explains how to build a team of GrokBot agents from scratch
worth more than a $500 agentic engineering course
skip Netflix today and watch this, then read how to build a fleet of GrokBot agents in the article below
$META unveiled Muse Charm, a palm-sized 5G device built specifically for talking to its Muse AI agent.
They plan to ship the device in December, in time for the holidays.