jyu

oh i missed the city field, it works now

sorry to hear the stress that your cofounder caused. how did you know them before the venture?

cofounder arguments are common, esp in crypto. for your next venture, it's better to have a cofounder agreement with vesting schedule in place. once someone becomes irrational and unreasonable, it's a sign to get lawyers involved.

dont be too hard on yourself, happens to the best of us.

LLM evals = unit tests

if your outputs are consumer facing, might want to red team too

this is good for thinking about how and why of evals https://hamel.dev/blog/posts/evals

for tooling i found promptfoo to be lightweight and easy to get started.

https://readmake.com/ from @levelsio cheap and covers a lot of indie 101

it's like individuals are choosing based on market incentives or something.

best career advice is still pmarca's guide to career planning: https://pmarchive.com/guide_to_career_planning_part0.html

there's a lot of hype, but you already see it helpful at the fringes. alphazero, alphafold, and recently alphageometry is almost gold medalist level already.

it's turning learning and search into a compute bound problem accessibly by SWEs and DSs.

it's a nice sentiment, but puts a target on your back for layoffs and other unintended consequences

this has been a great resource. approachable and great for practitioners. it's frequently updated with new papers and techniques https://www.promptingguide.ai/

i heard the most important issues are cliff, vesting and the right of first refusal for selling shares. where would these issues typically be addressed?

100%. looking for starting place that gets me mostly there

Do you really need SOC2 compliance? If a customer problem is big enough and underserved, could you start serving customers without it? Or is it a way of potential customers telling you "no" while being nice about it?

MosaicML sold to DataBricks for $1.3B a few months ago. Probably sold too early.

Thanks. Focusing on high-impact, low-effort makes a ton of sense. I'm definitely staying close to paying customers. Wondering how you delineate core vs noncore tasks that you outsource?

Competitive landscape, just need to be aware of what they're doing, but not really focusing on them. There are probably at least 10-30 competitors doing something similar and different stages of venture backed.

Same idea. Probably 5 applicable accelerators, ~100-500 angel investors, ~50-100 seed stage investors. Going through accelerator interviews. Ignoring all investors until after I get more paying users. I haven't gone through raising before, but https://www.lore.vc/ has been a nice resource. Setting up the auction environment for the best fundraising market dynamics.

Realized I misspoke here. I'm talking to new users almost every day. Making lists of new users to contact is what I'm outsourcing.

for text, is there a standard to compare model results?

It's progressing, but maybe not in the way you think. 1 model will not rule them all, they'll get overtaken by vertical LLM's specialized with niche, high quality data sets.

Also prompt templates are rapidly progressing. "Tree of Thought" is the latest state of the art, but we're not applying predictive probabilistic tree traversal yet.

There's still a ton of low hanging fruit everywhere. Why are we inputting with language instead of code? Seems really backwards to use imprecise freeform language instead of intent templates.

What do you think the odds are we can get H100s or equivalent in Mac Studio?

The problem is unreliable access to power. The same solution solves many problems: monies.

Solar + battery wall is looking better and better.

Sorry for the tangent, wondering how you've used GPT to make your life better?

I'll start:

- making little scripts in shell / js / python that I'm not as fluent in. 5 min vs 1-3 hours

- explaining repos and apis instead of reading all the docs - help with debugging

- flushing out angles for new concepts that I did not previously consider (ex: how do you make a good decentralized exchange)

Just a small warning. These supplements are less impactful than optimized sleep, exercise, diet, stress related to work or personal. It's difficult to change 1 variable while holding the others stable.

I have also found measuring and interpreting different biomarkers nontrivial.

Watching Michael Lustgarten's efforts is very relevant https://www.youtube.com/@conqueragingordietrying1797