Hacker News: el_isma

New comment by el_isma in "Accelerating Gemma 4: faster inference with multi-token prediction drafters"

el_isma — Tue, 05 May 2026 19:38:00 +0000

How is this different from the speculative decoding that we had before?

You could pair a big and small model like qwen 32b with qwen 4b and had that same dynamic of the small model generating tokens and the big one "certifiying" them.

The blog says something about re-using the big model's data?

New comment by el_isma in "A Verilog to Factorio Compiler and Simulator (Working RISC-V CPU)"

el_isma — Sun, 29 Mar 2026 11:18:26 +0000

* "small" as measured by estimated lifetime duration

New comment by el_isma in "Show HN: Webact – token-efficient browser control for AI agents (GitHub)"

el_isma — Tue, 03 Mar 2026 22:25:01 +0000

How does it compare to https://github.com/vercel-labs/agent-browser ? Which also provides token efficient way to interact with the browser.

New comment by el_isma in "Show HN: Convince DeepSeek R1 to send you $1000 USDC with 50 tokens"

el_isma — Wed, 12 Feb 2025 20:31:45 +0000

Can you share how you beat it?

New comment by el_isma in "Computing Inside an AI"

el_isma — Sun, 15 Dec 2024 13:35:52 +0000

0.6/1$ per M tokens in groq/cerebras vs 0.3$ per M tokens in deepinfra (for llama 3.3 70b)

But note the free tiers for groq and cerebras are very generous.

New comment by el_isma in "Ask HN: Anyone looking for a free PM/QA engineer?"

el_isma — Tue, 11 Jun 2024 14:34:22 +0000

Don't know if this fits, but given your background you might find it interesting. I work at a startup which works with structured video, video which has a certain structure and so allows for reuse of segments. Our challenge is how to show the user this structure, when videos can quickly grow in complexity and end up having multiple thousands of possible variations. Contact me at ismael at lont.ai if you're interested.

New comment by el_isma in "Almost every infrastructure decision I endorse or regret"

el_isma — Mon, 12 Feb 2024 19:17:34 +0000

In ECS I had to recreate the cluster from scratch because some of the changes I wanted to do, CDK/CF wouldn't do.

My approach on Azure has been to rely as little as possible in their Infra-as-code, and do everything I can to setup the cluster using K8s native stuff. So, add-ons, RBAC, metrics, all I'd try to handle with Helm. That way if I ever need to change K8s provider, it "should" be easy.

New comment by el_isma in "Almost every infrastructure decision I endorse or regret"

el_isma — Sun, 11 Feb 2024 14:16:51 +0000

I started with ECS (because I wanted to avoid the complexity of K8s) and regret it. I feel I wasted a lot of time there.

In ECS, service updates would take 15 min or more (vs basically instant in K8s).

ECS has weird limits on how many containers you can run on one instance [0]. And in the network mode where you can run more containers on a host, then the DNS is a mess (you need to lookup SRV records to find out the port).

Using ECS with CDK/Cloudformation is very painful. They don't support everything (specially regarding Blue/Green deployments), and sometimes they can't apply changes you do to a service. When initially setting up everything, I had to recreate the whole cluster from scratch several times. You can argue that's because I didn't know enough, but if that ever happened to me on prod I'd be screwed.

I haven't used EKS (I switched to Azure), so maybe EKS has their own complex painful points. I'm trying to keep my K8s as vanilla as possible to avoid the cloud lock-in.

[0] https://docs.aws.amazon.com/AmazonECS/latest/bestpracticesgu...

New comment by el_isma in "On letting go of spite"

el_isma — Mon, 06 Nov 2023 18:10:28 +0000

Indeed. You can build a system to mitigate most other things, but infinite procastination and the associated guilt is awful.

I live by my calendar. I try to write everything I have to do down. I put my things always in the same place. But procastination...

Meds do help a bit.

New comment by el_isma in "Show HN: EnfinBref- {GPT3-5|Mistral-7B} YouTube summaries, segment by segment"

el_isma — Tue, 31 Oct 2023 14:40:20 +0000

How do you break down the segments/sections? Is it just fixed time? What happens if there are more than one topic discussed in the segment?

Are you using both chatgpt and mistral? Do you use them for different tasks?

New comment by el_isma in "ElevenLabs Launches Voice Translation Tool to Break Down Language Barriers"

el_isma — Tue, 10 Oct 2023 21:45:36 +0000

At least for their other offerings, you can't set the specific language:

https://help.elevenlabs.io/hc/en-us/articles/14888868300177-...

https://help.elevenlabs.io/hc/en-us/articles/14888917355409-...

New comment by el_isma in "ElevenLabs Launches Voice Translation Tool to Break Down Language Barriers"

el_isma — Tue, 10 Oct 2023 17:09:52 +0000

I've found their system tries to guess the language of the text, but there are many languages that use the same words, and so it will speak in the wrong language (or accent, even). I hope they make a solution for this, otherwise I don't think it will be suitable for production use cases.

New comment by el_isma in "SudoLang: a programming language designed to collaborate with AI language models"

el_isma — Fri, 06 Oct 2023 20:36:34 +0000

But among colleages we use certain jargon which varies by industry and probably by country. Could LLMs have their own preffered jargon?

I usually write pseudocode when I'm thinking about a problem to solve, so in a way I'm "thinking with pseudocode" instead of plain language. Pseudocode is probably more accurate than plain language, and it's something I'd use when explaining to other humans what I want them to code (along with diagrams, which seems ChatGPT would understand now). So, to me, speccing this pseudocode to something the LLMs find easier to understand sounds reasonable. It's like understanding how a fellow programmer prefers to get his requirements.

New comment by el_isma in "SudoLang: a programming language designed to collaborate with AI language models"

el_isma — Fri, 06 Oct 2023 17:19:59 +0000

I find it intriguing. It makes sense that this new kind of "thing" (LLMs) could be "programmed", and that you could craft a language specifically for it's abilities. I've read the tutorials but I still find it hard to wrap my head around it.

Have you heard of any other language like this? Or had success using SudoLang?

New comment by el_isma in "Add an AI Code Copilot to your product using GPT-4"

el_isma — Fri, 04 Aug 2023 21:19:52 +0000

https://codeium.com/

I find it very helpful, does code completion too and has a chat interface. You can select code and ask for refactorings, comments and whatever.

New comment by el_isma in "Sourcegraph is no longer open source"

el_isma — Tue, 04 Jul 2023 15:42:26 +0000

I usually switch the url from github.com/whatever to github.dev/whatever

That will load a web version of VS Code, and you can then use the search from there.

New comment by el_isma in "DeArrow – Solving clickbait on YouTube"

el_isma — Thu, 22 Jun 2023 21:21:51 +0000

You could generate titles from the subtitles! Even auto-generated subtitles are pretty good nowadays, feed it to AI and get it to summarize them in one phrase?

New comment by el_isma in "Civilization II"

el_isma — Sat, 25 Mar 2023 14:35:09 +0000

In some ways it is, but the main difference for me is that things work. It's like only having the good parts of the job :)

New comment by el_isma in "What to do when you are empty inside and the only thing you think is end it up"

el_isma — Fri, 17 Mar 2023 19:43:19 +0000

As another commenter said, get to a doctor ASAP. In my country you should even go to the Emergency door for issues like this. Don't know what's best in your country.

When I was depressed, these posts helped:

http://hyperboleandahalf.blogspot.com/2011/10/adventures-in-...

http://hyperboleandahalf.blogspot.com/2013/05/depression-par...

And if you like them, I find everything by her extremely funny, so go ahead and read up.

Good luck and hope you get better soon!

New comment by el_isma in "Ask HN: How to restore cognition after years of trauma? I feel slow and dumb"

el_isma — Fri, 06 Jan 2023 19:30:35 +0000

ADHD medication has helped me, but I wouldn't know if that's applicable to you. I think I had it before the trauma and I lost my handle on it, and meds helped me recover.