Will there be room for your post?
In X's open code a new post now gets 2 hours to catch on, not 48, and a trial of 200 views, not 1000. My take: social networks are getting ready to bin tons of bot posts.
#AIAssistant #AIAgents
7
We’ve launched a new service focused on taking care of properties in Marbella: technical property management, repairs, diagnostics, coordination of specialist contractors, and preparing the property, garden and car before the owners arrive.
We also work weekends, so your plans can stay your plans.
More details: idgroup.es
8
X's feed parameters changed overnight: 48 hours became 2, and 1,000 views became 200. I've had my ups and downs, and here is what stayed with me: a platform's rules change, but your voice and the circle of people you help stay. The algorithm likes posts people like right away. And people like right away what they recognize as their own.
X ALGORITHM UPDATE 🚨
The fresh posts pool I told you about is now LIVE, and X rebuilt the small account boost on the same day (I check the open source algorithm daily with Claude and share updates worth sharing with you)
tl;dr:
- fresh posts pool: off → ON
- small account boost: up to 1,000 → up to 50,000 followers
- boost window: 48 hours → 2 hours
- boost view limit: 1,000 → 200 home views
- boost winner: highest score → best like rate
lesson learned: the first 2 hours of every post are now where it's won or lost, so post when your audience is actually online.
full details:
- The fresh posts pool is ON
Last time I said I'd tell you when it goes live. It's live.
Every time someone opens For You, X now pulls up to 200 posts from a pool of new stuff: under 8 likes, under 500 views, up to 2 hours old.
Your post still has to be good enough to get shown. But until now a brand new post with no likes had no lane of its own. Now it does.
- The small account boost now goes up to 50K followers
It used to stop at 1,000. Now any account up to 50,000 followers can get it.
It lifts one fresh original post to around 16th in someone's feed.
- But it only lasts 2 hours
The post has to be under 2 hours old (was 48) and have under 200 views from home feeds (was 1,000).
After that you're on your own.
- The winner is picked by like rate
Before, the highest scoring post won. Now X looks at how many of the people who've seen your post liked it, adds a bit of randomness so brand new posts still get a shot, and picks from the best two.
So the first people who see your post decide a lot.
- Under 1,000 followers? Mixed news
You still get the boost, but now you share it with much bigger accounts, and the window went from 2 days to 2 hours.
- Next thing to watch
X built a new way to find posts similar to what people recently engaged with. It's switched off. If it goes live, posts that stick to a clear topic win. I'll tell you.
18
Your customer did not ask a friend or Google. They asked their own AI assistant. A minute later they have one answer and a few links. If you are not in it, you lost and never found out. An AI assistant answers from what it found and understood. Two things matter: that you exist where it reads, and that your prices, hours, rules and common answers are described clearly enough for a machine. People call this generative engine optimization, but in plain words it means making sure AI finds you and recommends you. Start with what customers ask most and where a wrong answer costs most. That gives you a list, not a worry. I work through this with people one on one. Grok Bot from xAI, ChatGPT dots from OpenAI, Claude and Gemini Spark already work while a person is busy, and Meta says Muse is meant to become an autonomous assistant across its apps, Facebook and Instagram included. What will an AI assistant ask about you first? #AIAgents #AIAssistant #OpenAI #SmallBusiness #Entrepreneur
35
Yesterday I watched my AI agent, working from my inbox, trade messages with the support team of a service I use. The other side may have been an AI agent too. The two swapped neat emails with every detail needed, and the issue closed within an hour. It used to cost days and nerves. I do not think it is about typing speed: the agent has the whole history in front of it. So my idea is bigger than mail and messengers: one communication hub for messengers (more important than email today), email, a knowledge base, a personal database and cloud storage. I already see something similar in my work WhatsApp... Now I am thinking of gathering people who like building light processes without any coding, vibe coding included. With Ravivo, which we are shaping as a communication management service, content included, I want to show how easily what matters meets what you need: no soldering, plug in and the light is on. Curious? Drop me a message. #AIAgents #Productivity #WhatsApp #Claude #Ravivo
4
5
237
Every day brings another tool that already seems essential. While I am getting
comfortable with one, several more find their way onto the list.
It is easy to mistake the natural limit of our attention for falling behind.
But this feed has no final page. No one can finish it, ...
39
Peak season in real estate is not one screen. It is client chats, lawyer emails, contractors, viewings, new leads, developers and the context you still carry in your head. Ravivo is built around those live communications. #Ravivo #ChatGPT #Claude #RealEstate
16
I don't want another chat. The ones I work in are already open. This week the feed argues who wins: a separate agent app, the chat with a billion people, or the company that already holds your mail and calendar. I care whose letters and threads become the feed. People rebuild context every time they jump between ChatGPT and Claude. Keep files and rules with you, or the move costs more time than the agent. A mail AI already reads the inbox and hides the switch. For an agency and a restaurant, one daily job, two apps you already open and four fields you can check are enough. The bot drafts. Only what you will sign goes out. I do not hand over a client's life or a guest's life. Ravivo already runs the path of a post. We are adding connectors so the same path fits a message from an app you already use: the fragment is visible before it leaves, and the yes stays yours. This is an assembly, not a button that says every messenger is inside. #Ravivo #ChatGPT #Claude
35
When I coordinated lawyers on complex property deals, part of my job was to
keep the client's context intact across conversations. Not every detail they
trusted me with belonged in every conversation. ChatGPT brings the same
choice into a new setting: it can help with a narrow question very quickly,
but the temptation is to show it “the whole story” first.
That has changed how I think about Ravivo. We began with a clear content path:
prepare, check, approve, publish. For a real estate agent, the path before an
AI question matters just as much. Which part of a chat or document is actually
needed? Who allowed it to be shared? What stays inside the working conversation?
A content-management platform alone no longer feels as relevant as a service
for managing communications.
We are exploring connections between Ravivo, different channels and AI tools.
This is a direction of development, not a ready-made “ask anything safely”
button. The boundary is what interests me: an agent should be able to see
exactly what context leaves their workflow for the assistant they choose.
The AI gets a stronger question without receiving the client's entire private
life along with it.
#Ravivo #ChatGPT #RealEstateAI #PropTech
1
28
A paradox is taking shape around AI and small businesses. People can create
far more than before, and demand for these services is still high. Yet it is
getting harder to sell your work as an employee or contractor through the
constant noise claiming that “AI does everything for free, or at most $20 a
month.” In the eyes of employers and clients, that noise can devalue the
experience of people learning a new AI-related profession. Many small-business
owners start to believe they can do everything themselves because a chatbot
answers fluently. Others compare professional work with social posts, offer
absurdly low fees, or ask for a long free trial that should somehow be
customised for them in advance.
And there is another side to the paradox. The same belief slows the adoption
of genuinely useful AI services in small businesses: a fluent chatbot answer
is not a solution. Clients and providers are caught between what they want
and what each side can actually afford or deliver.
So continuous self-learning and an individual brand still seem to be among
the main sources of stability for people in the near future. All of us also
have to push through a growing wall of AI noise and choose a niche that is
truly our own, early enough to make it count.
#Ravivo #AI #SmallBusiness #AISkills #FutureOfWork
15
Recently I was debating with a friend how much trust you can place in regular ChatGPT conclusions when it gathers data through connectors and then analyses it on its own. The logic sounds convincing: there is a strong model, there is access to sources, there is a detailed instruction set, so a good result should appear on schedule. What caught my attention was something else: who checks the path itself. Did the model really open every source, collect everything it needed, and avoid losing part of the data along the way? At one point I pictured a brilliant office employee who forgets to request one piece of information, then puts music on and produces a polished report. That is where a convincing output stops being the same as a reliable process. #ChatGPT #AI #Automation
10
Fifteen years in real estate have shown me many ways a strong incoming lead can
go cold. One of the most common looks almost harmless: the enquiry arrives
somewhere other than where the agent is working at that moment.
One person sees the call. Someone else reads the WhatsApp message. A portal
enquiry lands in a shared inbox. The agent is at a viewing, on the road, or in
negotiations. For the team, these are several normal working channels. For the
client, it is one conversation that somehow keeps starting again.
A useful test for an agency: take the last ten new enquiries and check whether
three answers are visible in one place. Who owns the lead? What has already been
promised? When should the next step happen?
AI can capture the initial enquiry, preserve its source, prepare the context,
and remind the deadline. A named person still has to own the next step. The
client experiences continuity as service, not the number of systems behind it.
We applied the same principle in Ravivo to content operations. Ravivo is not a
real estate CRM. But an idea, its source, current version, approval, owner, and
next action should not dissolve between chats and tools. In the next posts, I
will show more of how that logic works inside the application.
#Ravivo #RealEstateAI #PropTech #AIWorkflow #CRM
8
I keep seeing larger custom skill libraries in the feed, sometimes with hundreds or thousands of entries. Useful as a library, yes, but the number alone guarantees nothing. In my case, one setup reached 167 skills and 75,807 characters of descriptions against a 30,000 budget. The model did not stop and say, “I could not read them all.” It simply reached the limit and carried on with whatever it had managed to see. With an index, the ceiling may come after the index and only some selected skills. Without one, it may stop somewhere near the first entries in sequence. Everything can still look normal, even with a decent result. My takeaway is simple: structure first, volume second. Split custom skills by project and check that the ones you actually need were loaded. #AI #AISkills #Automation
10
Recently I asked someone I know for a personal view on the UX / UI of a new project. I wanted his own reaction: what felt clear, where attention dropped, what created trust and what did not. He sent a few thoughts, followed by a ChatGPT analysis. And there was nothing odd about it. I use AI constantly myself. What stayed with me was something else. We increasingly form personal opinions after speaking with the same shared models, so the line between our own conclusion and an AI answer gets thinner. In business this is interesting: decisions are born in conversations where some thoughts have already passed through AI, while some uniquely human thoughts may never have appeared at all. It feels natural and useful. But where will the boundary of self identity be?.. #AI #ChatGPT #Business
24
In a demo, almost any service can look convincing, and that is where the trap begins. On screen, everything is tidy: the data arrives, each step behaves, and the result appears on time. In real work, though, value rarely shows up when everything goes right. It shows up when something goes wrong. Before paying for a new subscription, I find it more useful to damage the scenario on purpose: remove part of the data, imagine reversing an action, force a manual correction, or see what happens when an integration stays silent. Those moments reveal whether the product will support the business or add another layer of manual effort. They also show how quickly you can understand the failure and bring the process back under control. That is where SaaS maturity starts for me. #SaaS #AI #Automation
17
The most deceptive moment in no code, for me, does not arrive when something fails. It arrives the first time everything works. The form sends the data, the model replies, the CRM receives the result, and for a moment it feels as if the story is complete. That is usually where testing begins for me. I stop looking at the pretty path and start thinking about failure instead: who sees the error, who can stop the process, what stays in the history, and how control returns to a person if the chain moves in the wrong direction. Once you look at a build that way, the first success no longer feels like readiness. Speed matters, but confidence only appears when control, boundaries, and human ownership are clear in practice. That is when the first success feels believable. #nocode #AISkills #Automation #Codex #Claude
22
Grok bot feels like a rocket, but without structure you get rough edges…
Order before speed. Policies, disk map, sync — then tempo.
#AI #MultiAgent #Grokbot #FounderNotes
24
The most useful automation conversation I have starts oddly. Not with a
catalogue of models. With a request: show me last week’s enquiry that almost
died.
It is almost always the same story. A number was missing. The person could have
gone silent for three days. Someone still wrote, called, remembered. That is
the person being paid. The job title is secondary.
If that scene cannot be shown, a familiar purchase begins. A new chatbot. A
new vacancy. “AI for clients”. They are closing a hole they have not seen.
I learned this in expensive services, not on a diagram. A description almost
never holds interest. Memory of what is still missing does, until a decision
is made.
The useful takeaway can be used tomorrow. Before hiring and before buying AI,
name three things: which enquiry almost died, who did not let it go, and how
you will know it was not lost. That already shows whether to configure what
exists, buy, build, or leave the person alone for now.
#AI #SalesProcess #CustomerExperience #BusinessStrategy
20