Reward engineering isn’t about chasing a single metric. It’s about teaching your system what success really looks like.
Here’s how I’m approaching multidimensional evaluation, delayed feedback, and adaptive rewards:
bit.ly/47eKKbt

I'm super excited to have joined Zapier! So excited about the opportunity to help build the platform there!

I'm leaving a job that I love for an opportunity that I will also love. I've never had such conflicted feelings about leaving a company before. I've grown so much and am so grateful for the trust placed in me.

Just dropped part two of my series on probabilistic prompt pipelines:
Selector design for LLM systems.

Covers:
– Embedding-based retrieval
– Bandits (Thompson Sampling)
– Diversity-aware content
– Cold start solutions
– How we track what’s working right now

blog.alexmlewis.com/building-the...

Static prompts work until they don’t.

I wrote about how they break down in production and how probabilistic prompt selection keeps systems fresh, adaptive, and improving over time.

blog.alexmlewis.com/why-static-p...

I started using SuperClaude a few days ago and its been a really nice addition to my development workflow. Funny enough, I literally was in the process of adding really significant / commands to my claude code setup when I found it. github.com/NomenAK/Supe...

I think Claude Code is Anthropic's testing ground for AI agents, not a coding assistant. CLI filtered for technical users. That usage would give detailed usage for a more generalized release. More thoughts here: bit.ly/4kdfFZ1

Last couple of weeks of AI/LLM talks have been interesting in talk about AI system design. Some of that has led to discussion of "context engineering" being the new prompt engineering. @langchain.bsky.social has a great blog on it: blog.langchain.com/the-rise-of-...

The world of LinkedIn that I experience is possibly real people making AI hot takes (with generated content that’s not really meaningful) and other possibly real people replying to them with generated replies that’s not really meaningful 🫠

Being 4 months behind in ai/llm engineering feels like being 2 years behind in software engineering. My brain loves the pressure to keep up, my body not so much

There's almost no education resources I use year over year. @egghead.io has been one I've subbed to continuously since 2014. Thank you @joelhooks.com for putting so much into it. Thank you @johnlindquist.com for pushing the content boundaries of the platform. Ya'll have helped make my career

Just ported 6,000 lines of Python to TypeScript in a day using Claude Sonnet 4.

The trick? A simple markdown checklist that acts as "external memory" for the AI. Cut my sessions from 18 to 7 and reduced bugs by 70%.

Turns out structure beats cleverness.

bit.ly/4jWWB0U

I’ve been loving implementing some Gen ai features into the app I’m working on , but wow it’s hard to convince people that Gen ai features are really boring without focus and user data to make the Gen ai have meaning. If anyone has tips on getting buy in from product leaders, I’d love to hear it

95% of AI influencers I see are just marketing influencers that end up having you sign up for a newsletter. What a waste of air

Going to attempt to make a poolish based bread tonight/tomorrow. I’m really curious to see the flavor difference between it and the basic same day white bread I’ve made previously

Using XML for LLM prompts has been a game-changer. Defining <role>, <instructions>, and <examples> gives Claude 3 Haiku a structure it excels at following, improving accuracy and safety in my personal finance RAG system. This has been a system I’ve been working on for almost a year.

Baked my first bread today. It turned out pretty good (a bit dense in a couple of spots in the middle), but it gave me so much anxiety during the process for some reason. Can't wait to try again

I've always struggled with meeting notes and fearing I missed something

Now, I’ve automated the fear away:
🎤 Krisp for transcription
⚙️ Zapier to connect it all
📁 Google Drive for storage
📝 Notion to stay organized

Here’s how: bit.ly/42048qy

How do you handle meeting notes? Always open to ideas!