prash

I question pretty much everything.

[prash @heyprash]

building something for two people only (more for couples).

one vault. one key. one password.
no accounts. no emails. no tracking.
your memories, encrypted before they leave your device.

waitlist is open → https://t.co/Wj3IXT1bWQ

[Sherwood @shcallaway]

Ohhh so that's why they call it "Hacker News"

[Nick Khami @skeptrune]

society might genuinely collapse if tailscale gets hacked

[prash @heyprash]

It's time!

[prash @heyprash]

BM25 is the Wayyyy!

[Zed @zeddotdev]

We tested 25 approaches to MCP tool search, including decision models, LLMs, and embeddings. Our benchmark runs 150 search queries against 536 tools.

We settled on weighted BM25 with a few normalization rules. Fast, local, no model or API calls.

[prash @heyprash]

If you need a lot of skill files for the Harness or your model, ngmi!

[prash @heyprash]

You're not using FFmpeg enough!

[prash @heyprash]

Damn! This hits home!

[ian brash @iansbrash]

Being non-technical in San Francisco must feel like divine retribution

At least Fitness SF knows their customer base

[prash @heyprash]

>Found a OSS which is really cool
>had a bug affecting functionality
>got it fixed for myself
>now using it!

[prash @heyprash]

@kushikraj07 Yep! For the price, a Mac mini is great for local LLM inference, especially with unified memory. An M4 Pro with 24GB should be plenty for smaller quantized models. Go 48GB+ for larger ones. NVIDIA has better throughput, but gets expensive when you need more VRAM.

[prash @heyprash]

@SamsungMobileUS

[prash @heyprash]

@SamsungMobileUS Is a fucked up company. They are showing me the 3 prices for the same product and it changes in each step.

So on the product page it shows me a price and I put it on the cart, it's a different price. And when I reach checkout, it's a different price.

[prash @heyprash]

This is so cool. They give a reset and then you spend the reset in less than six hours. Such Efficient models

[prash @heyprash]

That's the most weird thing I've ever seen.

[Yum⋆₊˚ @yuhasbeentaken]

here are my 5 tips if you’re deciding between deepseek’s direct api and opencode go:

1. if i’m spending around $20–25/month on deepseek, i’d stay on the direct api. it gives me more control and i’m not locked into someone else’s limits.

2. if i’m coding 4–5+ hours a day, opencode go starts to make a lot more sense. that’s where the subscription economics can beat pay-as-you-go.

3. i’d still keep the direct api as a backup. i have some friends who are already using opencode as primary and switching back to api when it’s slow or unavailable.

4. for heavier workflows, i’d use opus for planning/review and run multiple deepseek agents in parallel for implementation. that’s probably the best price/performance setup right now.

5. my biggest concern is quality. i still have a suspicion the opencode version may be quantized, and if that’s true, i’d expect more hallucinations, weaker instruction-following and worse codebase understanding.

cheap is great...

but only if the model you’re getting is actually the same model.

[dax @thdxr]

did you use an agent or was it a hand job?

[prash @heyprash]

Wow, tells us somethin'

[prash @heyprash]

@rekram11 @opencode @thdxr I guess restarting service worked for me! Thanks

[prash @heyprash]

The new version of @opencode is broken @thdxr
2.0.24. IT's crashing and hanging!

[prash @heyprash]

@rekram11 @opencode @thdxr I was in a middle of the work so I just rollbacked to .22
Let me update again and try the commands dax suggested to see if it fixes the issue. The issue was slow startup, models hung up(all providers), esc to interrupt not working(had to click 10 times to stop)