Dion Almaer

Dion Almaer

@almaer.com

Bringing nuance to a knife fight 🔪, surprisingly older than @bgalbs Past: Ajaxian, Shopify, Google, Mozilla, webOS, and more!

6K followers2.8K following175 posts

Dion Almaer
@almaer.com

★ Stitching with the new Jules API The Jules team finished off another ship week with an initial version of their API, allowing you to integrate with your friendly asynchronous squid engineer. In this post I show how I use Stitch to design a native app that uses the new API...

View full thread
Dion Almaer
@almaer.com

Step 3: Coded by Jules Finally, it’s time to build. I setup a repo that contains my spec as a README, and the screens that I want to build which have been downloaded from Stitch.

Dion Almaer
@almaer.com

★ Stitching with the new Jules API The Jules team finished off another ship week with an initial version of their API, allowing you to integrate with your friendly asynchronous squid engineer. In this post I show how I use Stitch to design a native app that uses the new API...

View full thread
Dion Almaer
@almaer.com

Step 1: Starting with the Spec Captures what you are looking to build: it’s design and implementation. Sometimes I work with an AI to go deep on the definition up front, and other times I stay shallower and flush out the spec as I add more features.

Dion Almaer
@almaer.com

★ Stitching with the new Jules API The Jules team finished off another ship week with an initial version of their API, allowing you to integrate with your friendly asynchronous squid engineer. In this post I show how I use Stitch to design a native app that uses the new API...

Dion Almaer
@almaer.com

★ Pools of Extraction I’ve started to notice some patterns in how I work on software projects. I create throwaway projects to learn from, and large repos with a bunch of ideas in. I sometimes extract targeted repos from them…

Dion Almaer
@almaer.com

… Along with mini specs that contain relevant context, I am able to use a variety of models to get higher quality results. This short post shares some details. blog.almaer.com/pools-of-ext...

blog.almaer.comPools of Extraction: How I Hack on Software Projects with LLMsI’ve started to notice some patterns in how I work on software projects. Specs as mini-pools I often sit down and just write a spec. Even if I never build it, the spec itself is a kind of temporary…
Dion Almaer
@almaer.com

✨Stitch Design Variants: A Picture Really Is Worth a Thousand Words? We have some great Stitch updates for you today, & I posted about a couple. At first I talk about design variants and how to use the fact that your brain can grok a lot of images very quickly... blog.almaer.com/stitch-desig...

blog.almaer.comStitch Design Variants: A Picture Really Is Worth a Thousand Words?My head has been in the world of code generation for the last few years. Now that I’m also thinking about design generation, I’m seeing a lot of similarities but also some fascinating d…
Dion Almaer
@almaer.com

OH “We got too many Product Managers who were good at competitive analysis but not deeply technical, so we ended up with fear based copy cat roadmaps.”

Dion Almaer reposted

AI Native Dev
@ainativedev.io

JVM devs, this one’s for you: Rod Johnson (Spring) just dropped an OSS agent framework. JVM-native, Kotlin-based, and focused on agentic flows mixing LLM-prompted interactions with code and domain models. #Java #Kotlin #AINativeDev #News #OSS

Dion Almaer reposted

AI Native Dev
@ainativedev.io

One of the most interesting announcements at I/O got just a few seconds of airtime: Gemini Diffusion. A diffusion-based language model that: • Generates in parallel, not left-to-right • Hits ~1500 tokens/sec • Can self-correct mid-generation Read all about it here: https://www.ainativedev.co/99j

Dion Almaer
@almaer.com

Gemini Diffusion! I love a good zig and zag. We used to be in a world where image generation used diffusion techniques (from noise), and text used autoregresion (next token). ChatGPT added 4o image generation that went the other way and used autoregression …

Dion Almaer
@almaer.com

And today we see the zag. We have another tool in the toolbox, diffusion models for text. This isn’t new, and has been talked about for some time. Unlike next token, diffusion by its nature is a good fit for generating text in a non-sequential manner…

Dion Almaer
@almaer.com

If there are gaps that you want to fill in, this could have better results. It also allows for more parallelism, and doesn’t fall into the trap of being stuck on a path due to the words already generated. I’m keen to see how this works for codegen!!

Dion Almaer reposted

AI Native Dev
@ainativedev.io

AI tools multiply, and so do config files. “This has all given birth to a plethora of rules files that we need to manage.” Vibe-rules CLI discovers and syncs them so you can code not babysit .yaml. Details ➜ www.ainativedev.co/nbp #AINativeDev #DevTools #AIDevelopment #CLI

Dion Almaer
@almaer.com

% parade list openai I've been enjoying building simple tools writing a spec and having them generated. `parade` uses Ink and Vercel's AI SDK to let me explore model names from various providers (because I am using the openai:gpt-4o type URI format in other tools). github.com/dalmaer/parade

Dion Almaer
@almaer.com

Excited to launch the landscape and start to really curate the tools that devs should be aware of as we use AI to get better and building software!

AI Native Dev@ainativedev.io

Announcing the AI Native Dev Landscape: A place for developers to understand the AI tools that help them build software faster and better. Be the first to hear about new products and keep track of updates! ainativedev.co/yom

Dion Almaer
@almaer.com

I giggle whenever I see this in prompts: ``` IMPORTANT: Use valid markdown only for all your responses and DO NOT use HTML tags except for artifacts! ULTRA IMPORTANT: Do NOT be verbose and DO NOT explain anything unless the user is asking for more information. That is VERY important. ```

Dion Almaer
@almaer.com

Late binding of choices is a good thing and helps with choice fatigue. Convention over configuration helps here, and LLMs being able to “fill in the gaps” takes it to a whole new level. It favors getting things working and iterating from there in a tight loop.

Dion Almaer
@almaer.com

Voice is a great way to sculpt vs draw. You can be loose, go back and forth, and it isn’t about speaking the perfect thing which is what always held me back. Voice has become a real input for me these days. You?

Dion Almaer
@almaer.com

📽️ Podcast: The One with Two Guys talking about AI software engineering 📽️ As global as our aperture is there are constant reminders of how it's a small world. I got to co-host an episode of AI Native Development with two Israeli founders that I have and have had the pleasure of working for...

View full thread
Dion Almaer
@almaer.com

... Practically: how do you handle the fact that every engineer is working on their own branch with local changes? Please join me in your podcast viewing/listening application of choice in the comment below. I'm curious for your thoughts across these topics too! #ainativedev

Dion Almaer
@almaer.com

📽️ Podcast: The One with Two Guys talking about AI software engineering 📽️ As global as our aperture is there are constant reminders of how it's a small world. I got to co-host an episode of AI Native Development with two Israeli founders that I have and have had the pleasure of working for...

View full thread
Dion Almaer
@almaer.com

... applying systems that include LLMs to the practice of real world enterprise software. For example, many people talk about the importance of retrieval, but Augment has done the work to lead in this area. Context sure if king, but how do you make sure to get the right context?

Dion Almaer
@almaer.com

I also really enjoyed the discussion around fine-tuning and RAG. Once again the details matter. What does it mean to fine-tune a model on your codebase? What does fine-tuning do well in this situation, and where does it fall down? ...

Dion Almaer
@almaer.com

📽️ Podcast: The One with Two Guys talking about AI software engineering 📽️ As global as our aperture is there are constant reminders of how it's a small world. I got to co-host an episode of AI Native Development with two Israeli founders that I have and have had the pleasure of working for...

Dion Almaer
@almaer.com

Guy Gur-Ari has a deep research background, and has built a world class research team at Augment, which I saw first hand. The team is amazing at applying research, and in the episode you get to hear about a lot of the learnings that have come from...

Dion Almaer
@almaer.com

★ English will become the most popular development language in 6 years Based on @simonwillison.net's AI predictions from the @oxide.computer computer crews podcast, I have one for AI software development, and here I explain why and how. blog.almaer.com/english-will... #ainativedev

blog.almaer.comEnglish will become the most popular development language in 6 yearsSimon Willison wrote up his 1, 3, and 6 year AI predictions from his chat with the lovely Oxide crew. This is always a fun thing to ponder as you start a new year, especially since 2025 f…
Dion Almaer
@almaer.com

Had a fun trivia game that used our own answers to questions and I woke up so sad that one friend had “Microsoft Authenticator app” as their most used application 😱