Fabrice Gallet
@fbgallet
Followers
885
Following
3K
Media
134
Statuses
779
Leveraging and developing Tools for thought. TfF to make a better use of one's knowledge. Knowledge management to live a more fulfilling life.
Joined June 2011
I just released Philo🧩Craft,. a crafting game powered by AI (gpt-4o) to explore philosophical concepts, inspired by Infinite Craft by @nealagarwal !. Be the first to discover new concepts! .
3
1
15
Working on a smoother way to chat with results of Ask your graph Agent of Live AI in Roam (still a lot of work on UI, but the core capabilities are already there)
0
0
22
I'm completely revamping "Ask your graph" Agent from Live AI extension in @RoamResearch . The first version was a bit clunky. What I'm preparing will be much more satisfying!. E.g. it will test several strategies and semantic variations to discover the most relevant results.
1
0
16
So you could build an entire Roam graph on some topic with just a few prompts and querying it from another graph 🤔.
@RoamResearch MCP loop 🤯. Running roam-research-mcp by 2b3pro directly in Roam, creating some stuff in Roam via Live AI LangGraph agent.
2
2
13
Connecting to local and remote MCP servers directly in @RoamResearch, it's a bit tricky (need a local bridge for not HTTP native servers). but it's working very well !. Here is an example with @readwise local MCP server:
1
2
18
Testing to connect to a local MCP server directly from Roam, using Live AI and LangGraph for agentic use of the available tools. It works 👍 (tested with my own philosophy manager MCP).Not as smooth as in Claude Desktop but probably enough to make it worthwhile!.
0
2
14
I finally work seriously on Personal Philosophy Management (PPhiM), primarily based on discovering and examining one's fundamental beliefs + practical purpose. I'm currently developing and testing an MCP server and. wow, just mind-blowing! 🤯
2
0
9
Yeah, still a threat for critical thinking, and a good reminder that it's primarily our expectations about LLMs that are often unfounded!. We need to get used to using LLMs more wisely. It's the purpose of my Critical Reasoning Toolkit in Live AI for @RoamResearch
LLMs have revolutionized our interaction with artificial intelligence. However, beneath their impressive capabilities lies a concerning tendency: sycophancy—the inclination to prioritize user agreement over independent reasoning and accuracy.
0
1
10
In short: pure Vibe coding isn’t for me outside quick prototypes. But relying now and then on AI assistant really saves time and boosts motivation on truly creative parts. Here is the vibe coded tennis angles simulator: 🎾✨.6/6.
0
0
1
4) Here’s the catch: when AI does everything, you don’t learn much. Quickly reviewing code isn’t enough to grow your skills. Relying too much can actually dull your coding sharpness. ↓ 5/6.
1
0
0
3) On the bright side, Vibe coding excels at tackling tedious tasks, like refactoring. It’s powerful for those repetitive chores that can drain your energy if done manually. ↓ 4/6.
1
0
0
2) Yet, pure Vibe coding can feel a bit dull. The thrill of creating and problem-solving shrinks when the AI handles most of the work, even if the results are correct. The joy isn’t gone, just a bit muted. ↓ 3/6.
1
0
0
1) Like in learning tennis, patience is key. The AI often makes mistakes, sometimes frustratingly so. But with steady guidance, it always finds a way to fix problems. ↓ 2/6.
1
0
0
Just released my first (mostly) Vibe coding project: a tennis angle theory simulator using Cline+Claude Sonnet 4. 🧵 ↓ 1/6
1
1
3
New updade of Live AI is published! Including:.- easily define daily notes range as context.- support o3-pro, Sonnet 4, Grok 3 (with live search). - block sourcing (if you need to retrieve blocks in large context).
Coming soon: in the next version of Live AI for @RoamResearch it will be very easy to use Daily log as context, define a precise time interval and get references to source blocks. 📅🚀.For immediate cost control, an estimation of the context tokens and cost will appear instantly.
0
2
15
Coming soon: in the next version of Live AI for @RoamResearch it will be very easy to use Daily log as context, define a precise time interval and get references to source blocks. 📅🚀.For immediate cost control, an estimation of the context tokens and cost will appear instantly.
1
2
16
A big update of Live AI extension for @RoamResearch has been released !. Including:.- Image generation.- Mermaid & argument map generation.- Web search with OpenAI or Anthropic models. and a lot of other improvements and fixes.
3
5
38
Image generation/edition with OpenAI gpt-image-1 model in @RoamResearch soon available with Live AI extension next update !. Also Mermaid diagram generation, better :q queries from natural language, TTS and more.
2
3
20
@baibhavbista GPT-4.1 models will also be available in the coming update, but you can already add them manually in "Custom OpenAI models" setting: gpt-4.1, gpt-4.1-mini, gpt-4.1-nano.
0
1
5
By the way, I recommend reading @baibhavbista 's excellent article: whose arguments and analyses I fully agree with. And of course, I believe that an IA generated argument map is a tool for stimulating reflection but not a replacement for reading! ;-).
1
0
7