I used to be confused about when to use Skills vs code for AI agents
Now I think of it this way:
• Early stage → Skills. Optimize for iteration speed
• Production → Code. Optimize for cost, scale, and reliability
Let the LLM do the thinking. Let code do the orchestration
"Agents are just scripts."
That's what people say.
Then you build one for production.
Writing the script is easy. Maintaining it across countless edge cases is the hard part.
Miss one tiny condition or an important piece of context, and the entire execution can fail.
One thing that surprised me while evaluating LLMs:
The same step by step prompt that helped a smaller model hurt a stronger one.
The stronger model didn't need the extra scaffolding. I was in its way.
Prompt quality isn't universal.
That's why we need evals, not intuition.
Some things I've been doing lately to save on AI tokens:
• Caveman mode
• Better model for planning → cheaper model for coding
• Run /insights to spot token waste
• Create a project index/map for the codebase
• Start a new chat before context gets too big
@NTFabiano It’s been 4 months since I quit short-form videos on IG and YouTube. Less stress, more control, better life. Just stopping this one thing improved other areas of my life. Never want to go back to that again.
Watched a video yesterday where an engineer broke down how they decide whether to join a company. It felt way more like an investor mindset than a typical job search.
• Revenue velocity: is the market pulling the product or are they pushing it
• Market size: is this a space that can actually get big over time
• Customer obsession: do users love it enough to care if it disappears
• Moat: can this be wiped out by an AI wrapper or big tech in 6 months
• Transparency: are they honest about revenue, growth, and metrics
• Due diligence: what are real users saying, not just the company
• Learning density: will this role actually stretch you
• Risk vs reward: is the upside worth the uncertainty
Made me rethink how I look at opportunities
What do you look for before joining a company
1K Followers 1K FollowingLeading @DumpingLinks
I talk about software development, stocks, personal observation, and sometimes politics.
Views are personal
32 Followers 228 FollowingI build AI systems for businesses — DM for custom builds | AI Developer @SIZH_IT | Building PreClaw(Launching very soon, Follow ) | Google SC’25
#buildinpublic
634 Followers 325 FollowingFull-Stack + AI Engineer @ Techorigins.
Built 14+ SaaS, OSS author (cron-safe, AI-Sentry).
Now building Pannly — AI radar for product opportunities.
46 Followers 195 FollowingBuilding CCIP @chainlinklabs | Distributed Systems and Blockchain Engineer | prev. @AmazonAds , @Paytm | All about tech and career growth
448K Followers 6K FollowingChief Scientist, Google DeepMind & Google Research. Gemini Lead. Opinions stated here are my own, not those of Google. TensorFlow, MapReduce, Bigtable, ...
312K Followers 1K FollowingBuilding new things @thinkymachines. Also dabble in robotics at NYU. Cofounded @PyTorch. AI is delicious when it is accessible and open-source.
479K Followers 1K FollowingML/AI research engineer. Ex stats professor.
Author of "Build a Large Language Model From Scratch" (https://t.co/O8LAAMRzzW) & reasoning (https://t.co/5TueQKx2Fk)
1.2M Followers 787 FollowingProfessor at NYU & Executive Chairman at AMI Labs.
Ex-Chief AI Scientist at Meta.
Researcher in AI, Machine Learning, Robotics, etc.
ACM Turing Award Laureate.
1K Followers 484 FollowingOne API to scrape, enrich, and understand the web; powering 5,000+ companies with real-time context.
Backed by YCombinator (S26)
Run by @mynameisyahia