Hacker News Reader

mikejulietbravoon Apple is going to raise device prices, but when?

came to say this, there's almost zero history for this so unless their supply chain got meaningfully distorted I can't see why this would happen now

mikejulietbravoon The Traditional Vi

text editors 5ever

mikejulietbravoon How to get GLM 5.2 to 280 tokens per second

as benchmarked on AA

mikejulietbravoon Ask HN: How do you choose a model for a task?

The short answer is that it depends how well you define the boundaries of the task and the relative complexity. For example, smaller model is usually fine for something like summarization, but an "easier" coding task might still actually be quite difficult unless you eval it heavily like @paulhoule said

mikejulietbravoon The Math Behind TurboQuant

how many mentions of silicon valley?

mikejulietbravoon Nvidia Invests $150M in AI Inference Startup Baseten

seems like more than invested in other inference players

mikejulietbravoon After Babel Fish: The promise of cheap translations at the speed of the Web

nuances like idioms can and will be solved. wispr flow is already solving a lot of these things via their speech-to-text interface.

as better models are introduced, figurative language, implication, cultural nuancec etc. becomes easier to reconcile.

mikejulietbravoon Hyperion: Minecraft game engine for custom events

does the plugin architecture come with added overhead (assuming you add a lot of elements)?

mikejulietbravoon A brief history of threads and threading

was just glad this wasn't about twitter :)

mikejulietbravoon An analysis of DeepSeek's R1-Zero and R1

Awesome - we'd love to have our CEO/CTO chat with you and your team if you're interested. Shoot me a note at mike.bilodeau @ baseten.co and I'll make it happen!

mikejulietbravoon An analysis of DeepSeek's R1-Zero and R1

Mike from Baseten here

We're super proud to support this work. If you're thinking of running deepseek in production, give us a shout!

mikejulietbravoon Hundred Rabbits is a small collective exploring the failability of modern tech

The start of this reads like the beginning of a cult manifesto, but then transitions to a very logical solution for an important problem.

Baffling. I'm in

mikejulietbravoon Flux: Open-source text-to-image model with 12B parameters

What's the tl;dr on a difference from this to SD?

mikejulietbravoon Show HN: I built an indie, browser-based MMORPG

This is such a wild undertaking. I love it

mikejulietbravoon DoorDash and Pizza Arbitrage (2020)

This is the biggest point. You can simply not use them. Cook your own food or walk or drive or bike to a restaurant.

There is a massive market of lazy AF people who also are terrible at personal financial management, and they are the ones complaining.

mikejulietbravoon DoorDash and Pizza Arbitrage (2020)

Taxes are not the answer, let the free market figure it out

mikejulietbravoon Show HN: Dorkly – Open source feature flags

I'm kind of surprised they haven't already built this into the LD core

mikejulietbravoon Launch HN: Hatchet (YC W24) – Open-source task queue, now with a cloud version

I remember reading that post, there were a lot of good ideas in the comments

mikejulietbravoon Show HN: Baseten Chains – Framework and SDK for Multi-Model AI Products

The new docs are great. Awesome work there

mikejulietbravoon Personal Machines and Portable Worlds

Odd, this looks like it was posted 6 months ago but is back?

mikejulietbravoon Reversing for dummies – x86 assembly and C code (Beginner/ADHD friendly)

not a criticism here, just a genuine question. What about this makes it friendly for ADHD? Is it a formatting/style thing? Information ordering?

mikejulietbravoon Launch HN: Meticulate (YC W24) – LLM pipelines for business research

This was a huge problem when I was working on a similar product some years ago. The issue is that false positives or worse, false negatives, can be pretty catastrophic when you're thinking about valuing companies in particular. What's hard for LLMs (currently) is to decide what is most important or to infer the things that aren't explicitly stated.

mikejulietbravoon Launch HN: Meticulate (YC W24) – LLM pipelines for business research

This sounds a lot like what Grata does, although they've built a pretty impressive proprietary dataset

mikejulietbravoon Open Source Inference Engine Baseten Raises $40M from IVP, Spark and Greylock

I was surprised to see this wasn't already on HN given Sarah Guo is attached and the Truss project seems pretty cool

mikejulietbravoon The financial value of open source software

One of the problems I have with the analogy being made here is that most open-source projects were created because someone needed to solve a problem or do something to build something else. The other thing is that in software, more than anything else, it's usually okay to be imperfect or incomplete. An incomplete or imperfect drug is dangerous.

The other thing is that fundamentally OSS is a symbiotic relationship between creators, maintainers, and users. If I need to build a tool or tech for my own product, making it available to others gets me feedback, gets me ideas on how to improve it, hardens it, etc.

mikejulietbravoon Tuxedo Computers

Name something hacker news readers like more than a literal linux box

mikejulietbravoon Amazon Just Got Banned from the EU Parliament

https://commission.europa.eu/about-european-commission/servi...

There's an entire thing. Not an expert by any means, but I think it would be sort of enforceable/distinguishable

mikejulietbravoon Amazon Just Got Banned from the EU Parliament

I think the distinction here might be that lobbyists often work for outside firms and in some countries may require to be registered as such.

mikejulietbravoon Show HN: CriticalPath – Advanced Profiler for Android

This is super cool! I know it's super new, but are there plans to support more languages in the future? Groovy would be nice

mikejulietbravoon I am the OP of that $104K bill post and I have some follow-up questions

This story is absolutely bananas