As the release of DeepSeek and other open-weight LLMs resets expectations around model performance and infrastructure demands, we’re witnessing a shift in how capital is being allocated in the AI arms race. The assumption that running powerful models requires a warehouse of GPUs is starting to crack—and that changes everything. (FT article on capex shifts, Investors.com coverage)
At What Comes Next, LLC, we’re committed to building ethical, transparent, and accessible AI systems—not just in philosophy, but in practice. As part of the R&D process behind The Catalyst (our AI-enhanced behavioral coaching platform), we are documenting what it actually takes to run modern LLMs locally, on consumer-grade hardware, from a real-world testbed.
This whitepaper will map out that journey—not in theory, but through direct experimentation. Over the next several weeks, we’ll be:
- Building and tuning a testbed with a desktop development machine and a local server in a residential environment
- Addressing real-world constraints like physical security, LAN segmentation, and network isolation without enterprise budgets
- Running DeepSeek and other open models using quantization and lightweight deployment layers
- Benchmarking performance, heat, power draw, and token throughput across local configurations
- Outlining the post-MVP architecture for scale without relying on cloud dependency
This isn’t just an experiment—it’s an attempt to lower the barrier for others. If we want AI to be ethically deployed, then it must also be accessible. Our systems should not require loyalty to a tech giant or blind trust in black-box APIs.
By releasing this whitepaper and the accompanying tools and documentation, we aim to provide a clear, replicable path for solopreneurs, researchers, and early-stage builders to run LLMs at the edge—securely, affordably, and transparently.
Check back soon for the first release.
In the meantime, follow along with the build threads, benchmarks, and previews on our LinkedIn and GitHub pages.
Because if AI is going to shape the future, we all deserve a seat at the table.