Gabe De Mesa

Software Engineer

[Gabe De Mesa @jamesjellow]

My dad decided to come back to his province in the Philippines after leaving to America for a better life. This is what happened when he came back:

[Gabe De Mesa @jamesjellow]

@elonmusk Currency would need to evolve

[Gabe De Mesa @jamesjellow]

@theo so far very methodical, especially when reviewing pull requests. feels like a bump in quality, and in cost.

[Gabe De Mesa @jamesjellow]

@theo How has reliability been with Bun? We're thinking of migrating services over but I am concerned about it not as battle tested as Node.

[Gabe De Mesa @jamesjellow]

Some photos from #GTC2026 !

[Peter Steinberger 🦞 @steipete]

We should make EnterpriseClaw just for the lolz.

Java 21, Spring Boot, 14 abstract factory beans, 2GB Docker image, takes 45 seconds to start, AbstractSingletonProxyFactoryAgentClawResponseHandlerBeanDefinitionRegistryPostProcessorImpl .java

[Theo - t3.gg @theo]

AI might actually take my job…

[Gabe De Mesa @jamesjellow]

Best game I’ve ever watched in my life #SNF #joshallen #bills

[Gabe De Mesa @jamesjellow]

Cross posting a message I sent to my company's AI guild:

Can AI think?

This question is sparking debate in the AI community because folks are arguing that AI models cannot "think" for itself similar to that of a human. On one hand we have Apple, where they argue that Large Reasoning Models (LRMs) "does not provide insights into the reasoning traces’ structure and quality."  I've heard folks on this end of the spectrum also argue that LLMs are just next token predictors on their training set/context. https://t.co/Y0v2NBF8zV

On the other hand, we have the leading model labs like OpenAI's o3/o4-mini; and Anthropic Claude 4 Opus which highlight impressive AI performance on difficult benchmarks and argue these models do reason, just not exactly like humans. When prompted differently, models can solve complex tasks more reliably. But is it truly reasoning or is it simply memorizing and regurgitating answers? (o3/o4-mini system card: https://t.co/3wGXWDPwPA, claude 4 opus system card: https://t.co/Rdg7xmc88r)

The disagreement boils down to how we define “thinking”: which in my opinion is the innate human-like capability to learn understand, and breakdown problems. But is it truly thinking similar to that of a sentient human being or just acting as a glorified search engine? Implicit biases and company interests also shape these views: Apple tends to emphasize caution and downplay the capabilities of AI since they notoriously have been held accountable for their recent shortcomings with Siri, whilst model providers highlight the potential however many have a direct interest in selling said models as a service.

I don't think I'm qualified enough to say if an AI can truly reason or think similar to that of a human, however I think we can observe that it exhibits certain qualities and perspectives of a thinking machine (See LLMs pass the turing test: https://t.co/ZtUxrY0FV1).

It's important also to note that certain tasks will always be inherently human such as emotional intelligence or tasks that require elevated levels of human touch or care albeit major investment and R&D is being put into this field (see Claude article on EQ: https://t.co/D5Gx1eovFg).

My conclusion I sometimes have to remind myself that AI is inherently going to be better for solving NP/NP-complete problems especially if it evolves remembering every minute details within a data corpus to arrive at an answer especially since there's no way a human could store and search hundreds of thousands of tokens of context at once.

#ml #turingtest #ai #llm #openai #anthropic #apple

[Dhruv Bindra ⏳ @bindra_dhruv]

Open request to the memory team @OpenAI

Responses are annoyingly biased by what I’ve spoken about before (for eg a fresh discussion on a startup idea will suddenly respond with an India focus)

I find myself either using temporary chat (which sucks because I want to be able to go back to it) or switching to other LLMs

Would love a toggle to opt out of “Use Memory” in stored chats

[Gabe De Mesa @jamesjellow]

Jennie’s full #likeJENNIE performance! 🔥 #BLACKPINK_DEADLINE #BLACKPINK #JENNIE

[Gabe De Mesa @jamesjellow]

Bruno Mars just pulled up on night 2 LA #BLACKPINK #BrunoMars #rose #apt

[Gabe De Mesa @jamesjellow]

World's First Autonomous Delivery of a Car | Tesla https://t.co/gYUaAb57b9 via @YouTube

[Gabe De Mesa @jamesjellow]

@ProductHunt @cursor_ai @AnthropicAI Claude code integrates directly into Cursor. So I'd say both. If I had to choose one, Claude Code for sure.

[Gabe De Mesa @jamesjellow]

@disarto_max Bottom for sure. I feel it's more compact and easier on the eyes I'd even go as far as stacking the "14" bubble on top of "selected" to follow the design pattern.

[Gabe De Mesa @jamesjellow]

@niccruzpatane I would get so dizzy 🤣

[Gabe De Mesa @jamesjellow]

@Starlink Here's my 2cents. Starlink is a great option definitely for those who live in rural areas. I wonder if starlink ever plans to create their own ISP so those of us in suburban or urban areas can hardwire in? @Starlink just giving you guys some ideas lol.

[Gabe De Mesa @jamesjellow]

@mattiapomelli @cursor_ai @figma DM Sent!

[Gabe De Mesa @jamesjellow]

Sharing @karpathy's YC keynote. He's an incredible communicator, and visionary for AI. #yc #ai #llm https://t.co/R5521fwjAT