themanmaran

8 ms·

91.
by themanmaran 1y ago
This is a big problem when it comes to conversational agents. Sometimes users ask questions that are really prying, potentially misleading, or just annoying repeats (like asking for a cheaper price 50 times). In these situations a real pers
92.
Weightlifting tracker for coders(github.com)
1 points by themanmaran 1y ago|0 comments
93.
by themanmaran 1y ago
I think the negativity is overkill. I've been toying around with the exact same idea the last couple weeks. It's mostly GPT 4o image => some image cleanup, but honestly a lot more finicky than I originally expected. Lots of pro
94.
by themanmaran 1y ago
Apparently the new number is $1500. But it's at the tow driver's discretion so they can estimate the value at $1,499.
95.
by themanmaran 1y ago
Got curious about the "century old statute" and looked it up. Original text [1] > Section 1. Whenever any motor-vehicle is found abandoned upon any public highway… and said vehicle is not of a value exceeding one hundred dollar
96.
by themanmaran 1y ago
Yea but for Mistral it's not just the enterprise plan. It's all variants of the "Mistral Code". So pretty useless to launch on HN if the only CTA is booking a demo.
97.
by themanmaran 1y ago
The original source has much better historical information and a definition of the "functional unemployment" rate [0]. And the charts here don't really show this rate "soaring" > the percentage of the U.S. labor
98.
by themanmaran 1y ago
Why is that? AWS has availability zones all around the globe, and 8 across the EU.
99.
by themanmaran 1y ago
The post below got flagged, but they the link they shared has some really nice visualizations of the scaffolding: https://multimedia.scmp.com/infographics/culture/article/318...
100.
Show HN: Extension that swaps "AI" for "army of interns"(github.com)
2 points by themanmaran 1y ago|0 comments
101.
by themanmaran 1y ago
The user patterns are a bit different in each case. When I'm copy-pasting from GPT, I'm forced to define the problem better and be specific on the output I'm looking for. Compared to Cursor with full codebase context where I
102.
by themanmaran 1y ago
Yea the founders also have majority common stock. So there's not a normal scenario where the founders and other investors get paid out in an exit, but the angels don't. The bigger fear is a non-exit scenario, where the company bec
103.
by themanmaran 1y ago
Airbyte is probably the best opensource tool in this space.
104.
by themanmaran 1y ago
Nice and simple! I'm excited for all the fun micro businesses that get enabled by the new image API. Things like your coloring book, instant sticker/tshirt/swag creation, video game assets, etc. Also love the "tap 5 time
105.
by themanmaran 1y ago
This is the new model that's available in ChatGPT, which most notably can do transfer generation. i.e. "take this image and restyle it to look like X". Or "take this sneaker and give me a billboard ad for it"
106.
by themanmaran 1y ago
Likely because they've seen a lot of the potential abuse capabilities. i.e. the "generate a drivers license with this face". So the options are: 1) nerf the model so it can't produce images like that, or 2) use some type
107.
OpenAI releases image generation in the API(openai.com)
487 points by themanmaran 1y ago|297 comments
108.
by themanmaran 1y ago
I think it's less malicious, and more generally tech debt. Gmail is incredibly intertwined with the world. Around 2 billion daily active users. Which makes it nearly impossible for them to ship new features that aren't minor tack
109.
by themanmaran 1y ago
True, probably better generalized as "recency advantage". A startup like Brex has a huge leg up on traditional banks when it comes to operational efficiency. And 99% of that is pre-ai. Just making online banking a first class expe
110.
by themanmaran 1y ago
The horseless carriage analogy holds true for a lot of the corporate glue type AI rollouts as well. It's layering AI into an existing workflow (and often saving a bit of time) but when you pull on the thread you fine more and more reas
111.
by themanmaran 2y ago
How are those mutually exclusive statements? You can't imagine someone working on backend (focused on distributed systems) for 10-15 years at a FANG company. And also being in a position to interview new candidates?
112.
by themanmaran 2y ago
Using robo-taxies to pass notes like a message in a bottle. Not the future I expected, but honestly sounds like a fun little experiment.
113.
by themanmaran 2y ago
I disagree. From the average user perspective, it's quite confusing to see half a dozen models to choose from in the UI. In an ideal world, ChatGPT would just abstract away the decision. So I don't need to be an expert in the rela
114.
by themanmaran 2y ago
I'm curious when it crossed the line into "fraud" here. Since almost every "AI" application has tons of human fallback. Waymo has human drivers that can teleoperate the vehicle when it gets stuck. The Amazon Go stor
115.
by themanmaran 2y ago
I see a fair bit of utility here. We have an API product, and (despite having pretty clear docs) I almost always have to send people curl requests they can copy and run. So I see this as similar to having a sandbox built into your docs page
116.
by themanmaran 2y ago
> Never change the original language of any text. Keep Korean in Korean, Japanese in Japanese, and English in English. I love the double prompting to keep GPT from translating the text. I've definitely had this problem before, and s
117.
by themanmaran 2y ago
How does queue observability work in hatchet? I've used pg as a queueing system before, and that was one of my favorite aspects. Just run a few SQL queries to have a dashboard for latency/throughput/etc. But that requires you
118.
by themanmaran 2y ago
We're also looking to test qwen and other for the bounding box support. Simon Willison had a great demo page where he used Gemini 2.5 to draw bounding boxes, and the results were pretty impressive. It would probably be pretty easy to d
119.
by themanmaran 2y ago
That was the cost when we ran Llama 90b using TogetherAI. But it's quite hard to standardize, since it depends a lot on who is hosting the model (i.e. together, openrouter, grok, etc.) I think in order to run a proper cost comparison,
120.
by themanmaran 2y ago
Generally running the whole benchmark is ~$200, since all the providers cost money. But if anyone wants to specifically benchmark Omni just drop us a note and we'll make the credits available.

More ›