Comments URL: https://news.ycombinator.com/item?id=50262244

Points: 2

# Comments: 0

Yeah, having review/orchestration + subagents seems like the stndard way to optimize for cost/performnce.

Using OpenCode has been on the list to try. Logically, the value of a bundle from a larger company should technically beat out a smaller purveyor of open-weight models right? I also wonder how much OpenCode's token/sec throttling will impact my productivity?

This is why I'm comparing value for the money.

While Google may not have the "best" models, what's the point if you hit a wall with Claude Code? The extra baseline quota with Anti-gravity could be worth it.

I currently have access to ChatGPT Pro, GitHub Copilot, etc. I've used Claude Code, but can't really distinguish the vibes between the models. However, I often run out of tokens at the "Plus" plan levels even when using GPT-6.1-Sol as my daily driver.

I'm considering Google Antigravity Ultra at the $100 tier. It also seems to include a generous rate limits and quota for their models and series of extra features including: cloud storage, YouTube Premium, etc.

I figure that if the Google models don't perform as well as the "frontier" labs then I can always supplement with $20 plans and have the Anti-gravity tooling use the more capable models as orchestrators and use subagents.

What do folks think, what's the best value for the money bundle out there right now?

Comments URL: https://news.ycombinator.com/item?id=50167220

Points: 4

# Comments: 4

If you're worried about costs, you might try using subagents with a cheaper model or an open model like Kimi/GLM and use Claude/Codex as the orchestrator. This is very efficient.

I increasingly find myself wanting to go back to being customer-facing. I'm not super interested in this whole FDE thingy. It seems like it's going to fizzle out once VC subsidization of consulting services expenses becomes less en vogue.

Is Sales Engineering still a fun job? What's the market like right now for more Senior/Staff-level SE?

Comments URL: https://news.ycombinator.com/item?id=49527390

Points: 2

# Comments: 0

> Why do I want to make music?

I picked a basic DJ controller and a midi controller bundled with Ableton. I'm a novice, but I love listening to music and dissecting what makes a good performance. I crave that feeling of getting chills when I find something new that moves me in new ways. This set was a pretty recent example:

https://www.youtube.com/watch?v=gfF8jzBVWvM

That being said, the world is increasingly crowded with "good enough" music.

I resolved early on that I was never going to make a money doing this, which simplified things greatly. There's a primal part of our brain that craves adoration. I do wish for others to adore my music. Even if it's a handful of people. I do wish to perform publicly one day, even if it's at a park for passersby.

Mostly I just want something to move my brain in different ways. I want to create something beautiful.

> ...and noticed companies have become less likely to offer their time for ride-alongs and research calls. They get too many requests, and vibe coding is drawing their attention to self-build.

Is this ACTUALLY happening? Are entrepreneurs who get into vibe-coders really eating up time a bunch of time for trades people?

So they may not be willing (even though they are able) perform procedure/test if they aren't confident they'll get paid.

No doubt it's astounding, and I have no experience underwriting these kinds of scale of investments...

...but if I were a recent employee diluted as part of this raise (even with the massive uplift in revenue) I'd be very skeptical about an sort of financial outcome for myself.

If I were an employee looking to join Entire, or a developer evaluating the durability of Entire for my needs over the long-term, I'd ask things like —

- What's the terminal value of a DevTool in the AI era? Is it closer to a standard 10x ARR? or maybe 100x...perhaps 1000x?

- Is there value in the asset attributable to the network? If so, is it defensible? What is the likelihood protocols emerge that simply disintermediate the network moat of a AI agent memory company like Entire?

- What kind of AI data are developers willing to silo with a single vendor? What kind of controls do Enterprises expect of an AI agent memory system? Can Entire reasonably provide them to grow in the next 12-24 months?

- As a potential employee...if you join a company with a $60M seed funding and 19 employees, what is the ARR they need to achieve based on the first product in market in roughly ~12 months? $6M...$12M...20M? Is there a DevTools company that's ever done that!? What is the thesis for "this time is different"? Getting developers to pay for things is tough 'doncha know?

Only then can you ask basic technical diligence questions —

- Is Git a system that can scale to be the memory system and handle the kind of tasks that AI agents are expected handle?

- Are the infrastructure costs for these early platform decisions here reasonable over the long-term? Or do they just run out of money?

I wish them the best, and I hope employees get liquidity, and people take money off the table in a responsible way.