Built Permesso Digitale with Francesco and Carlo: an AI agent that guides you through the nightmare of opening a restaurant in Rome. Permits, inspections, business plans, all in one place.
Your phone can now run AI without internet and it's completely free.
Google just pushed a big update to an app called AI Edge Gallery. It's been around since last year, but now it got Gemma 4 (Google's newest open-source model) and iOS support for the first time.
Here's the deal: You download an AI model to your phone (about 2.5GB) Once it's downloaded, everything runs offline No data leaves your device โ full privacy It's free: no subscription, no API costs
What you can do with it: - Chat with AI (similar to ChatGPT) - Ask questions about photos - Transcribe audio in real time - Run multi-step tasks with the new "Agent Skills" feature
Quick reality check: this isn't going to replace cloud AI like ChatGPT or Claude. Those are still more powerful for serious work. But here's why I think you should try it anyway:
It's free, so no more paying $20/month for basic AI tasks like transcription or quick drafts Your data never leaves your device: if you handle client information, that matters On-device AI is the future! Google, Apple, and Qualcomm are all betting on it. The models will get better fast, and the people who start experimenting now will have a head start
Simon Willison (one of the most respected voices in the developer community) called it "terrible name, really great app." Hugging Face's co-founder called the open licensing "a huge milestone." It already hit #8 on the App Store.
I'm going to be recording full tutorials on how to set this up and get the most out of it, so make sure you're subscribed.
We hosted our first AI & automation meetup at Hexa in Paris, an evening of real talk with business owners, executives, and operators who are done with the hype.
Three speakers, practical lessons, and a networking session fueled by good Champagne. We covered everything from modernizing legacy systems to building workflows that actually work.
Key takeaway? Most companies don't have an AI problem, they have a process problem. Fix the foundations first, then automate.
Big thanks to Lovable for hooking up the audience with discount codes so people can start building right away.
This was just #1. More coming soon: where should we go next? Drop your city in the comments.
Lovable is completely free right now. For the next 24 hours, the AI app builder that normally costs $25+/month is open to everyone โ no credit card, no trial, full access. The setup:
It's International Women's Day, and Lovable's SheBuilds initiative is making the entire platform free Anthropic is sponsoring it โ every participant gets $100 in Claude API credits Stripe is throwing in $250 in fee credits on top 30+ in-person events across 17 countries, plus a live Discord track
If you've never tried vibe coding, this is the easiest entry point you'll get. You describe what you want in plain English, Lovable generates a working web app. No coding needed. Previous SheBuilds participants have shipped real products that are still live โ AI Recess, MicroFundHer, Sawa. The only catch: it's 24 hours. After today, you need a paid plan to keep editing. Register at lovable.dev/invite/2GPWDND and build something. Even if it's small. Even if it's ugly. Ship it.
OpenAI just dropped GPT-5.4 โ and they did NOT slow down. Two days. That's the gap between GPT-5.3 Instant and this new model. Here's what you need to know:
33% fewer factual errors compared to GPT-5.2 Uses up to 47% fewer tokens to solve problems (faster responses, lower costs) You can now interrupt it mid-thought โ it shows its plan upfront and lets you redirect before it finishes 1 million token context window in the API (now officially out of beta) First OpenAI model with native computer use โ AI agents can now click, type, and navigate apps New ChatGPT integration that works directly inside Excel and Google Sheets Outperformed human professionals 83% of the time across 44 job categories
API pricing did go up though โ $2.50/$15 per million tokens vs $1.75/$14 for GPT-5.2. OpenAI says the efficiency gains make up for it. Available now for Plus, Team, and Pro users as GPT-5.4 Thinking. Free users get limited auto-routed access. The timing is interesting โ OpenAI is dealing with the Pentagon deal controversy, the QuitGPT movement picking up steam, and Claude/Gemini gaining ground. This feels like their response: ship faster, ship better. Full breakdown coming in a video soon. In the meantime โ have you tried it yet?
Google just launched Nano Banana 2 โ their new default AI image model, and it's a solid upgrade. Here's the quick breakdown: ๐ It merges the quality of Nano Banana Pro with the speed of Gemini Flash
Generates images 3-5x faster than Pro Gets about 95% of Pro's image quality Text rendering is significantly better โ readable mockups and graphics Style transfer โ apply any reference photo's aesthetic to your images Subject consistency across multiple characters Can use Google Search data while generating (knows real places, brands, etc.) Now powers image suggestions inside Google Ads campaigns Free in the Gemini app โ about 15-30 generations per day
Also: you can now upload any image to Gemini and ask if it was AI-generated. Google's SynthID watermark checker will tell you. Pretty useful. Available now in Gemini, Google Search (141 new countries + 8 new languages), Flow, and developer APIs. Paid plans start at $19.99/mo for higher resolution and more generations. I'll be posting prompts and real examples tomorrow so you can see the actual quality difference between NB2, Pro, and the original. Stay tuned for that. If you've already tried it โ what do you think so far?
OpenAI just released GPT-5.3 Instant โ here's what actually changed: The big theme: ChatGPT was too "cringe." Overbearing tone, unnecessary refusals, dramatic preambles before answering simple questions. OpenAI says this update fixes that. Key improvements:
More natural, direct conversational tone 26.8% fewer hallucinations when using web search Fewer unnecessary refusals on questions it should answer Better contextualized web results (not just link dumps) Less moralizing and defensive phrasing
This isn't a new model generation โ it's a polish update to the everyday ChatGPT experience. GPT-5.2 Instant stays under Legacy Models until June 3. Available now for all users. API: gpt-5.3-chat-latest Thoughts? Drop them below ๐
๐ Claude just launched "Import Memory" โ transfer your ChatGPT context to Claude in under 60 seconds Anthropic dropped a feature that makes switching from ChatGPT (or any AI) ridiculously simple. How it works:
Go to claude.com/import-memory Copy the prompt they provide Paste it into ChatGPT โ it exports all your memories Paste the output into Claude's memory settings
That's it. Claude merges it with its memory system and you're good to go. Key details:
Works with ChatGPT, Gemini, and any AI with memory Available on Pro, Max, Team & Enterprise plans Memory is encrypted, editable, and exportable It merges โ doesn't overwrite existing Claude memories
If you've been on the fence about trying Claude, the barrier just disappeared. Full blog post on my site (link in description) ๐ What do you think โ game changer or gimmick? Drop your take below.
๐ OpenAI just unleashed GPT-5, free for every ChatGPT user! โฃ โก Super fast, smarter, with the lowest hallucination rate yet. โฃ ๐ง Dual engine: instant replies + deep-thinking mode, routed in real time for expert answers. โฃ ๐ Outperforms earlier models in coding, math, writing, health, vision, and more. โฃ ๐ฎ Build full games, ๐ craft pro dashboards, ๐ค tackle complex health questions in minutes, not weeks. โฃ ๐ Voice chats more natural. ๐จ Pick your chat color and persona to match your style. โฃ ๐ Mini models kick in after limits, and a unified mega-model is coming soon. โฃ ๐ Do you want to stay up to date on the latest AI developments, learn how to use AI in your business, and stay ahead of the competition? Follow @LivingOffAI for daily insights, tools, and breakthroughs!
Living Off AI
Officially a โช@lovableโฌ Ambassador! ๐งก
Been using Lovable to build and ship real full-stack apps every day for quite some time now.
What I love most about Lovable is how it enables anyone to build anything in a matter of hours, not months.
Now I get to share it with the community here!
1 month ago | [YT] | 3
View 0 replies
Living Off AI
We just won the Agentic Hack Rome at Urbe Hub !
Built Permesso Digitale with Francesco and Carlo: an AI agent that guides you through the nightmare of opening a restaurant in Rome. Permits, inspections, business plans, all in one place.
Powered by โช@lovableโฌ !
1 month ago | [YT] | 2
View 0 replies
Living Off AI
Your phone can now run AI without internet and it's completely free.
Google just pushed a big update to an app called AI Edge Gallery. It's been around since last year, but now it got Gemma 4 (Google's newest open-source model) and iOS support for the first time.
Here's the deal:
You download an AI model to your phone (about 2.5GB)
Once it's downloaded, everything runs offline
No data leaves your device โ full privacy
It's free: no subscription, no API costs
What you can do with it:
- Chat with AI (similar to ChatGPT)
- Ask questions about photos
- Transcribe audio in real time
- Run multi-step tasks with the new "Agent Skills" feature
Quick reality check: this isn't going to replace cloud AI like ChatGPT or Claude. Those are still more powerful for serious work. But here's why I think you should try it anyway:
It's free, so no more paying $20/month for basic AI tasks like transcription or quick drafts
Your data never leaves your device: if you handle client information, that matters
On-device AI is the future! Google, Apple, and Qualcomm are all betting on it. The models will get better fast, and the people who start experimenting now will have a head start
Simon Willison (one of the most respected voices in the developer community) called it "terrible name, really great app." Hugging Face's co-founder called the open licensing "a huge milestone." It already hit #8 on the App Store.
I'm going to be recording full tutorials on how to set this up and get the most out of it, so make sure you're subscribed.
Works on Android 12+ and iOS 17+.
1 month ago | [YT] | 0
View 0 replies
Living Off AI
AutomaTalks #1: Done. C'est fait!
We hosted our first AI & automation meetup at Hexa in Paris, an evening of real talk with business owners, executives, and operators who are done with the hype.
Three speakers, practical lessons, and a networking session fueled by good Champagne. We covered everything from modernizing legacy systems to building workflows that actually work.
Key takeaway? Most companies don't have an AI problem, they have a process problem. Fix the foundations first, then automate.
Big thanks to Lovable for hooking up the audience with discount codes so people can start building right away.
This was just #1. More coming soon: where should we go next? Drop your city in the comments.
1 month ago | [YT] | 2
View 0 replies
Living Off AI
Lovable is completely free right now.
For the next 24 hours, the AI app builder that normally costs $25+/month is open to everyone โ no credit card, no trial, full access.
The setup:
It's International Women's Day, and Lovable's SheBuilds initiative is making the entire platform free
Anthropic is sponsoring it โ every participant gets $100 in Claude API credits
Stripe is throwing in $250 in fee credits on top
30+ in-person events across 17 countries, plus a live Discord track
If you've never tried vibe coding, this is the easiest entry point you'll get. You describe what you want in plain English, Lovable generates a working web app. No coding needed.
Previous SheBuilds participants have shipped real products that are still live โ AI Recess, MicroFundHer, Sawa.
The only catch: it's 24 hours. After today, you need a paid plan to keep editing.
Register at lovable.dev/invite/2GPWDND and build something. Even if it's small. Even if it's ugly. Ship it.
2 months ago | [YT] | 1
View 0 replies
Living Off AI
OpenAI just dropped GPT-5.4 โ and they did NOT slow down.
Two days. That's the gap between GPT-5.3 Instant and this new model. Here's what you need to know:
33% fewer factual errors compared to GPT-5.2
Uses up to 47% fewer tokens to solve problems (faster responses, lower costs)
You can now interrupt it mid-thought โ it shows its plan upfront and lets you redirect before it finishes
1 million token context window in the API (now officially out of beta)
First OpenAI model with native computer use โ AI agents can now click, type, and navigate apps
New ChatGPT integration that works directly inside Excel and Google Sheets
Outperformed human professionals 83% of the time across 44 job categories
API pricing did go up though โ $2.50/$15 per million tokens vs $1.75/$14 for GPT-5.2. OpenAI says the efficiency gains make up for it.
Available now for Plus, Team, and Pro users as GPT-5.4 Thinking. Free users get limited auto-routed access.
The timing is interesting โ OpenAI is dealing with the Pentagon deal controversy, the QuitGPT movement picking up steam, and Claude/Gemini gaining ground. This feels like their response: ship faster, ship better.
Full breakdown coming in a video soon. In the meantime โ have you tried it yet?
2 months ago | [YT] | 1
View 0 replies
Living Off AI
Google just launched Nano Banana 2 โ their new default AI image model, and it's a solid upgrade.
Here's the quick breakdown:
๐ It merges the quality of Nano Banana Pro with the speed of Gemini Flash
Generates images 3-5x faster than Pro
Gets about 95% of Pro's image quality
Text rendering is significantly better โ readable mockups and graphics
Style transfer โ apply any reference photo's aesthetic to your images
Subject consistency across multiple characters
Can use Google Search data while generating (knows real places, brands, etc.)
Now powers image suggestions inside Google Ads campaigns
Free in the Gemini app โ about 15-30 generations per day
Also: you can now upload any image to Gemini and ask if it was AI-generated. Google's SynthID watermark checker will tell you. Pretty useful.
Available now in Gemini, Google Search (141 new countries + 8 new languages), Flow, and developer APIs.
Paid plans start at $19.99/mo for higher resolution and more generations.
I'll be posting prompts and real examples tomorrow so you can see the actual quality difference between NB2, Pro, and the original. Stay tuned for that.
If you've already tried it โ what do you think so far?
2 months ago | [YT] | 1
View 0 replies
Living Off AI
OpenAI just released GPT-5.3 Instant โ here's what actually changed:
The big theme: ChatGPT was too "cringe." Overbearing tone, unnecessary refusals, dramatic preambles before answering simple questions. OpenAI says this update fixes that.
Key improvements:
More natural, direct conversational tone
26.8% fewer hallucinations when using web search
Fewer unnecessary refusals on questions it should answer
Better contextualized web results (not just link dumps)
Less moralizing and defensive phrasing
This isn't a new model generation โ it's a polish update to the everyday ChatGPT experience. GPT-5.2 Instant stays under Legacy Models until June 3.
Available now for all users. API: gpt-5.3-chat-latest
Thoughts? Drop them below ๐
2 months ago | [YT] | 1
View 0 replies
Living Off AI
๐ Claude just launched "Import Memory" โ transfer your ChatGPT context to Claude in under 60 seconds
Anthropic dropped a feature that makes switching from ChatGPT (or any AI) ridiculously simple.
How it works:
Go to claude.com/import-memory
Copy the prompt they provide
Paste it into ChatGPT โ it exports all your memories
Paste the output into Claude's memory settings
That's it. Claude merges it with its memory system and you're good to go.
Key details:
Works with ChatGPT, Gemini, and any AI with memory
Available on Pro, Max, Team & Enterprise plans
Memory is encrypted, editable, and exportable
It merges โ doesn't overwrite existing Claude memories
If you've been on the fence about trying Claude, the barrier just disappeared.
Full blog post on my site (link in description) ๐
What do you think โ game changer or gimmick? Drop your take below.
2 months ago | [YT] | 1
View 0 replies
Living Off AI
๐ OpenAI just unleashed GPT-5, free for every ChatGPT user!
โฃ
โก Super fast, smarter, with the lowest hallucination rate yet.
โฃ
๐ง Dual engine: instant replies + deep-thinking mode, routed in real time for expert answers.
โฃ
๐ Outperforms earlier models in coding, math, writing, health, vision, and more.
โฃ
๐ฎ Build full games, ๐ craft pro dashboards, ๐ค tackle complex health questions in minutes, not weeks.
โฃ
๐ Voice chats more natural.
๐จ Pick your chat color and persona to match your style.
โฃ
๐ Mini models kick in after limits, and a unified mega-model is coming soon.
โฃ
๐ Do you want to stay up to date on the latest AI developments, learn how to use AI in your business, and stay ahead of the competition? Follow @LivingOffAI for daily insights, tools, and breakthroughs!
Source: @openai
#AI #Tech #News #OpenAI #ChatGPT
9 months ago | [YT] | 1
View 0 replies
Load more