whinvik

7 ms·

61.
by whinvik 10mo ago
Ok, I was a bit addicted to Opus 4.5 and was starting to feel like there's nothing like it. Turns out Gemini 3 Flash is pretty close. The Gemini CLI is not as good but the model more than makes up for it. The weird part is Gemini 3 Pro
62.
by whinvik 10mo ago
Interesting. Have you seen any benefits of using io-uring. It seems io-uring is constatly talked about but no one seems to be really using it in anger.
63.
by whinvik 10mo ago
Is there no transcript?
64.
by whinvik 10mo ago
The actual lesson I learnt from this post is that Python was good enough for Reddit until 2024 from 2005.
65.
by whinvik 11mo ago
I think what is happening is that people are realizing AI is not just plug and play. It can do amazing things but needs engineering around it. I think what will happen is in parallel more products will be built that address the engineering
66.
by whinvik 11mo ago
Sorry, I am going to ask this again. This is exactly how a DAG works? So is the hype around things like Durable Execution, Temporal etc simply because something like Airflow is not more well known?
67.
by whinvik 11mo ago
Sorry for the off-topic but I have been lately seeing a lot of hype around durable execution. I still cannot figure out how this is any different than launching a workflow in something like Airflow. Is the novel thing here that it can be do
68.
by whinvik 11mo ago
Haha I don't know if its a good test or not but I could not figure out why git pull was failing and Claude just went crazy trying so many random things. Gemini 3 Pro after 3 random things announced Github was the issue.
69.
by whinvik 11mo ago
I find the following way of expressing args awkward. with engine: fn(args) I would rather have with engine: fn(args, engine) This makes testing way easier.
70.
by whinvik 11mo ago
I feel finding 1 persona makes sense since otherwise you will have a messy time trying to build too many things. However, I also think selecting that single persona too early will hamper you more than messy building will. So its a balance.
71.
by whinvik 1y ago
What kind of usecases does a screenshot CLI have?
72.
by whinvik 1y ago
Anybody know what is meant by 'injecting relevant metadata'. Where is it injected?
73.
by whinvik 1y ago
I feel like this is how normal work is. When I have to figure out how to use a new app/api etc, I go through an initial period where I am just clicking around, shouting in the ether etc until I get the hang of it. And then the third o
74.
by whinvik 1y ago
My general experience has been that Gemini is pretty bad at tool calling. The recent Gemini 2.5 Flash release actually fixed some of those issues but this one is Gemini 2.5 Pro with no indication about tool calling improvements.
75.
by whinvik 1y ago
Just to clarify, my original comment was neutral about whether its a good or a bad thing. It was just a statement of observation.
76.
by whinvik 1y ago
Ads. They created ads. Now (or eventually) they can charge app developers to be featured first for a specific use case.
77.
by whinvik 1y ago
Yes, it feels very similar to Devcontainers. But having tried Devcontainers a lot, it feels like its probably a complex specification since outside of VSCode every implementation has issues. The CLI on Mac has weird user permission issues.
78.
by whinvik 1y ago
Same. I didn't really understand what the difference is compared to containerization
79.
by whinvik 1y ago
Having done some tests, its clearly better at instruction following and JSON output now. However its hampered by max output tokens. Gemini is at 65 K while GPT 5 mini is at 128K. Both of them have similar costs as well so as such apart from
80.
by whinvik 1y ago
Thanks. If I am interpreting this correctly, what you have is not a browser but a translation layer. You are still using something that scrapes the data and then you translate it to be in the format that works best for your agent. My origin
81.
by whinvik 1y ago
Nice, this looks interesting. > It uses a text-based browser we built Can you tell us more about this. How does it work?
82.
by whinvik 1y ago
What is the advantage over Duckdb with Spatial Extension.
83.
by whinvik 1y ago
Its odd. People on HN routinely complain how Stripe or PayPal or some other entity banned them unfairly and the overwhelming sentiment is that it was indeed unfair. But when it comes to this thread, the sentiment mostly is banning is good a
84.
by whinvik 1y ago
Have you tried using git worktree. Your tasks should be fairly independent though.
85.
by whinvik 1y ago
Threads like these are always one sided with people against RTO. What I would really like to see is arguments from the other side. Can someone steelman RTO. Preferably with evidence, anecdotal or otherwise.
86.
by whinvik 1y ago
The failed Windsurf bid and this makes me think OpenAI feels they need to focus more on the coding agent use case. Still thinking about the endgame. Its not obvious to me if OpenAI/Anthropic will become competitors to coding startups l
87.
by whinvik 1y ago
Is this exercise done to determine what the model can produce from its training data or is the data shown again to the model?
88.
by whinvik 1y ago
I really don't think this is a good idea. All the negative comments seem to have been from people who almost treated 4o as a friend rather than a tool. I don't think encouraging that direction is good in any way.
89.
by whinvik 1y ago
Curious. Are there real world usecases where people have finetuned such tiny models and put them into production.
90.
by whinvik 1y ago
I agree that FastAPI is not the best but I would not jump to another framework just because of an article. For FastAPI I know that I can go to [1] and see what a good FastAPI code base looks like. These days even Airflow is on FastAPI but h

More ›