Brian Love
@blovedev
Followers
2K
Following
3K
Media
275
Statuses
3K
Christian, Father of Evelyn, Wife to Bonnie, @GoogleDevExpert, building https://t.co/9gUkbiLKBW at @liveloveappdev, he/him.
Bend, OR
Joined September 2008
what a joy to hang out and build an MCP + generative UI app with my friend @MikeRyanDev on the web dev challenge! want to see what we built? hint, we built it with https://t.co/zHhMXnQz8A follow @codetv_dev so you can watch when it drops
1
1
6
this. this is why you won’t find pictures of my daughter anywhere online. the more difficult aspect of this decision was asking (and to some extent enforcing) that the g-parents don’t post to fb pics with her in it.
This is why you don't share just everything on the web. The vid you see here must be one of the most effective AI awareness videos.
1
0
1
LLMs are autocomplete for intent. As a UI engineer, that changes everything. Anywhere you can predict or complete what a user is doing, AI belongs there.
1
1
0
JSON is token‑expensive for LLMs – just like @mattpocockuk frequently mentions. Meet TOON, the Token‑Oriented Object Notation. 💸 40–60% fewer tokens than JSON 📐 readable & tokenizer-aware Wrap your JSON with `encode` to save half the token cost: https://t.co/UoG9yHmgfg
github.com
🎒 Token-Oriented Object Notation (TOON) – JSON for LLM prompts at half the tokens. Official spec & reference implementation. - johannschopplich/toon
210
347
4K
Just saw something from the actual future! The @LiveLoveAppDev team showed off #hashbrown and wow!! @MikeRyanDev and @blovedev showed some amazing GenAI demos.
2
2
5
untether us!
Codex team is cooking 🧑🍳 Now we need a really nice Codex cloud mobile app...
0
0
0
im here for the tech. this is cool. im interested to see how this plays out with non-tech folks. i also (kinda) enjoy novelty. i also know that novelty wears off. is this novelty? the market will tell us. i get fb + flash vibes.
Can ChatGPT run Doom? Yes. ChatGPT Apps are very powerful. I cloned our Next.js ChatGPT template, registered a 𝚙𝚕𝚊𝚢_𝚍𝚘𝚘𝚖 MCP tool and deployed to @vercel. Once the tool is called, ChatGPT embeds the full @nextjs application. Server and client rendering just works, and
0
0
0
very interesting talk from @aiDotEngineer conf on reinforcement learning my take away is that RL could provide gains in model improvements for tasks that can be automatically verified (like coding)
1
0
0
This week we held our biggest DevDay yet. And honestly we couldn't have done it without Codex! I went ahead and collected a couple of examples of how Codex helped us across the entire show https://t.co/eW8yu5yFvf
developers.openai.com
Learn how Codex helped us build experiences, demos, products, and more
17
40
293
You can also check out hashbrown. a client first alternative with hooks for streaming text, markdown, and even full UIs
1
0
0
agreed - just like you don’t want to useEffect with a fetch and ddos yourself.
been using vercel's ai sdk streamText() handles all the streaming stuff so you don't have to write fetch wrappers and stream parsers yourself makes it way easier to get ai responses flowing to the frontend
1
0
0
I was waiting for this…
OpenAI just launched AgentKit. Here's what it means for AI agent development, and an honest comparison with Langflow (we built the same agent in both) 👇 🧵 (1/8)
0
1
2
interesting, I also found this blog post that provided some more details
this was helpful! i built a search feature that uses gpt-5-nano to show search results using generative ui with hashbrown, but I was providing the data using json.stringify()
1
0
0
maybe this answers my question
Exciting news! We're working with OpenAI to make Apps mcp-ui compatible (and vice-versa). Both share same vision and philosophies - The future experience of the web is not one of distinct, complex websites anymore - but of fragmented, custom "UI chunks" composed by an assistant.
0
0
1