blog

insights gamedev tools view all posts

tools

Sift reset agitodo triday

Insights Gained

Sharing my latest learnings across LLMs, productivity systems, personal development, and the journey of building in public. These posts explore insights from my experience as a solo creator.

From my latest experiments with LLMs to learning about wellness to redefining productivity, learn along with me.

Structuring QwQ Reasoning Outputs

An approach to standardizing outputs from reasoning models via an instruction model harness.

Structuring QwQ Reasoning Outputs

Deepseek R1 32B + Qwen2.5 32B = ?

Testing how well a local reasoning LLM and a local instruction LLM can work together.

Deepseek R1 32B + Qwen2.5 32B = ?

LLM-powered brainstorming

Looking for value in LLM generations by prioritizing breadth of ideas over depth.

LLM-powered brainstorming

parental structures designed by LLMs

Seeing if a creator-critic architecture can design better frameworks for child development.

parental structures designed by LLMs

tasking LLMs to help the environment

Can the dual critic agent architecture arrive at impactful, plausible, environmental solutions?

tasking LLMs to help the environment

testing dual-critic agent systems

Evaluating a dual-critic agent architecture powered by Deepseek R1 vs. Dolphin R1.

testing dual-critic agent systems

testing actor-critic agent systems

A look at the results of an actor-critic agent architecture powered by popular models.

testing actor-critic agent systems

adversarial agents with a judge added

How adding a judge into adversarial systems affects performance using Deepseek-R1 32B.

adversarial agents with a judge added

game design via adversarial agents

Tasking agents to compete for who can come up with the best game design.

game design via adversarial agents

creative thinking in LLMs

In search of deeper value and leveraging reasoning patterns in Deepseek-R1 32B.

creative thinking in LLMs

o3-mini vs. its predecessors

Initial comparisons of o3-mini, o3-mini-high, o1, Sonnet 3.5, and Deekseek-R1 models.

o3-mini vs. its predecessors

competitive agent loops in Deepseek

Exploring the capabilities of competitive agent loops using Deepseek-R1 32B.

competitive agent loops in Deepseek

making wellness accessible

Exploring how to make mental health and wellness tools more accessible to everyone in 2025.

making wellness accessible

SEO keyword research tutorial

Learn how to find the right keywords for your SEO strategy in 2025.

SEO keyword research tutorial

casino-grade UI/UX design

A practice in learning more complex UI/UX design via a baccarat simulation.

casino-grade UI/UX design

why productivity is so attractive

The common obsessions with productivity and how to harness it effectively.

why productivity is so attractive

vitamin into painkiller

Making a case that sometimes a vitamin product can eventually become a painkiller.

vitamin into painkiller

why should I journal?

A look at the power of personal journaling and how it can help you grow.

why should I journal?

always redefining productivity

A look at how my perspective on productivity has evolved over time.

always redefining productivity

the right mentor at the right time

Thoughts on the importance of finding the right mentor so they're there when you need them most.

the right mentor at the right time

reading books for growth

A list of books recently read that have helped me grow as a person and developer.

reading books for growth

managing is freaking hard

Thoughts on the complexities of modern management.

managing is freaking hard

co-founder search difficulties

Sharing some of the difficulties I've encountered while searching for a co-founder.

co-founder search difficulties

refining my focus

A look at the guiding principles that are helping me focus on my goals.

refining my focus

welcome to KokuTech!

Gotta start somewhere.

welcome to KokuTech!