David Antunes @_DAntunes_

"If you want to build a ship, don’t drum up the men to gather wood, divide the work, and give orders. Instead, teach them to yearn for the vast and endless sea" antunesindustries.com Joined May 2012

Tweets

2K
Followers

198
Following

1K
Likes

35K

David Antunes @_DAntunes_

2 days ago

underrated use-case: use codex (gpt5) to brainstorm about new features, refactoring, etc.

0 0 1 24 0

David Antunes @_DAntunes_

5 days ago

my current loop: - write with gpt-5/gemini - polish and generate .md of feature request - run codex/gemini cli implement .md - ask the LLM if what shipped matches the intent

0 0 0 84 0

There’s a part in the book where he mentions that Shenzhen used to be known for its oysters, but industrialization wiped them out. Can’t help but think that in the US we would have picked protecting the oysters over the industrialization.

gabe @allgarbled

2 weeks ago

9 8 195 109K 128

Download Image

48 73 2K 116K 441

David Antunes @_DAntunes_

2 weeks ago

Gemini CLI in my current workflow: 1. creating detailed documentation md files. 2. use md files to brainstorm next steps (Gemini or chatGPT) 3. then implement with Gemini or chatGPT 4. rinse and repeat

0 0 0 46 0

David Antunes @_DAntunes_

2 weeks ago

Getting into manufacturing? Understand Monozukuri (at its core, not just the translation).

0 0 0 19 0

David Antunes @_DAntunes_

2 weeks ago

ok, I was to harsh to GPT-5, chatGPT is back being my go-to AI

0 0 0 29 0

Atoms Not Bits @AtomsNotBits

3 weeks ago

Google is backing the training of 100,000 new electricians to meet data center demand. Skilled trades such as HVAC and welding form the backbone of America’s industrial economy and its reindustrialization. @get_gild is the platform enabling the growth and support of the skilled…

16 31 222 21K 91

Download Video

David Antunes @_DAntunes_

3 weeks ago

🏭 Machine Spirit update: multi-factory(&line) planning It now can read live (simulated) factory data, and plan the optimal multi-factory production schedule. It then generates a full summary and a Gantt chart, turning raw optimization data into actionable strategy.

0 0 2 70 0

Download Video

David Antunes @_DAntunes_

3 weeks ago

Been thinking about this for a while, we might not need giant LLMs. Small models, connected with the right tools, can show emergent behavior. That might be the better option. Have been building in that direction. More on this soon.

0 0 0 56 0

David Antunes @_DAntunes_

3 weeks ago

The Factory Machine Spirit update time Our LLM agent can now execute multi-step plans, chaining tools like optimization and simulation in sequence. It then analyzes the combined results to provide actionable insights. This is the foundation for a fully autonomous factory.

0 0 0 42 0

Download Video

David Antunes @_DAntunes_

3 weeks ago

Are LLMs trained on our typo-ridden prompts? Are we teaching them to write like us… or worse?

0 0 0 19 0

David Antunes @_DAntunes_

4 weeks ago

Think beyond the demo. "Machine Spirit" is the factory’s nervous system in the making: visualize now, operate next, continuous agent ahead. below current tools, list will grow

0 0 0 30 0

Download Video

David Antunes @_DAntunes_

4 weeks ago

Sneak peek: labor added

0 0 1 19 0

Download Video

David Antunes @_DAntunes_

4 weeks ago

Ok, I have a new workflow using both Gemini Chat and Gemini Agent. Agent = developer. Chat = generates PRs and reviews the code, checking if it meets the goals and code review criteria. Both have full access to the repo.

0 0 0 23 0

David Antunes @_DAntunes_

4 weeks ago

Rare earths are not rare. The real bottleneck is refining. EU/France and the US are now moving to bring refining onshore. Plan for abundance, not scarcity.

0 0 1 49 0

Download Image

Aleksander Holynski @holynski_

4 weeks ago

Something fun we discovered: you can use #Genie3 to step into and explore your favorite paintings. Here's a short visit to Edward Hopper's "Nighthawks".

1K 1K 12K 9.6M 4K

Download Video

David Antunes @_DAntunes_

4 weeks ago

Updated the Factory Machine Spirit UI! Our digital twin now spins up multiple parallel instances, each with a configurable failure rate (placeholder values for now). This lays the groundwork for automatic schedule-recovery models and AI training.