HackerTrans
TopNewTrendsCommentsPastAskShowJobs

eevmanu

no profile record

comments

eevmanu
·19 giorni fa·discuss
Fugu Ultra <https://console.sakana.ai/models#fugu-ultra> sounds similar to GPT-5.5 Pro or Gemini 3.1 Deep Think .

Is there any official source that could confirms if Fable (or Mythos) is parallelized test-time compute (like GPT 5.5 Pro) or sparse Mixture-of-Experts (MoE) transformer combined with a multi-agent, inference-time compute scaling architecture (Gemini 3.1 Deep Think)?
eevmanu
·19 giorni fa·discuss
Reminds me of <https://github.com/irthomasthomas/llm-consortium>
eevmanu
·2 mesi fa·discuss
are you familiar with worktree? or are you asking for something more specific or elaborated?
eevmanu
·4 mesi fa·discuss
or https://github.com/adamritter/fastgron
eevmanu
·4 mesi fa·discuss
made me remember this article

<https://adamdrake.com/command-line-tools-can-be-235x-faster-...>

  Command-line Tools can be 235x Faster than your Hadoop Cluster (2014)

  Conclusion: Hopefully this has illustrated some points about using and abusing tools like Hadoop for data processing tasks that can better be accomplished on a single machine with simple shell commands and tools.
eevmanu
·4 mesi fa·discuss
Would you mind sharing or pointing me to any links that explain how you've set up Zig to work effectively with LLMs using agentic features?

I was thinking that downloading the full official documentation, separated by sections inside the repository someone is working on with Zig, could be useful, but maybe there are more optimal ways to approach this.

Thanks.
eevmanu
·6 mesi fa·discuss
..: Instant QR-based Account-to-Account (A2A) Payment System
eevmanu
·6 mesi fa·discuss
Six years old might be too early to incentivize screen use, though I'd encourage you to research this yourself rather than taking internet comments at face value.

That said, I think there's a distinction between screens and computing itself. You could introduce her to computing power through voice interfaces: a smart speaker connected to an LLM could let her search, learn, and interact with information hands-free. You'd have control over the system prompts for safety, and could whitelist reliable sources for her queries.

Yes, visual information density is higher than audio, but the downsides of early screen exposure might outweigh that efficiency gain. Voice-first computing could be a middle ground, she gets to explore what computers can do without the attention/addiction patterns that screens introduce.

Just one perspective, obviously. Worth doing your own research on the developmental tradeoffs.
eevmanu
·6 mesi fa·discuss
When I have that need, this is what I do.

I use the info from here as context:

https://github.com/InteractionDesignFoundation/add-event-to-...

I load it via a custom system prompt or skill, then pass my specific need, thought, or anything I want to remember in the future, including events with a certain frequency, as part of the prompt.

From that, I render the URLs needed to create the events directly in my personal calendar via the browser. This part could probably be automated better, but honestly I'm lazy sometimes.

And that's it.

I also try not to overdo it with high-frequency reminders, since that tends to de-incentivize actually using your personal calendar, which kind of defeats the purpose.

On top of that, Telegram 'Saved Messages' with the reminder feature is really useful. The native app makes it very fast to navigate. Obviously not as fast as searching local plain text files, but definitely faster than WhatsApp.
eevmanu
·6 mesi fa·discuss
Is that actually required?

I'm not at my computer right now, but as far as I remember most LinkedIn posts are viewable without logging in if you use the direct URLs, for example:

https://www.linkedin.com/feed/update/urn:li:ugcPost:<post_id>

https://www.linkedin.com/feed/update/urn:li:activity:<activity_id>

Unless something changed recently, those links should work anonymously.
eevmanu
·6 mesi fa·discuss
recently ...

1000x: The Power of an Interface for Performance

https://www.youtube.com/watch?v=yKgfk8lTQuE

by Joran Dirk Greef
eevmanu
·6 mesi fa·discuss
I have a similar workflow and I'd like to know how sophisticate is your requirement for annotations, because I solve that part kind of smoothly with Excalidraw
eevmanu
·6 mesi fa·discuss


  cr=country<two letter>
I have a "kind" of similar need and use *cr* query param based on *ISO 3166-1 alpha-2* [1] to force official language when i want to narrow the search on a specific country (as example when I want to search for english, I use *cr=countryUS*)

[1] https://en.wikipedia.org/wiki/ISO_3166-1_alpha-2

p.s.: YMMV, sometimes works, sometimes it doesn't, most of the time works but there is no determinism
eevmanu
·7 mesi fa·discuss
just out of curiosity, what is the difference between this and dspy optimizers?
eevmanu
·7 mesi fa·discuss
I’m a bit confused because I don’t clearly understand the value this tool adds. Could you help me understand it?

From what I can see, if the content I want to enrich is static, the web fetch tool seems sufficient. Is this tool capable of extracting information from dynamic websites or sites behind login walls, or is it essentially the same as a web fetch tool that only works with static pages?
eevmanu
·7 mesi fa·discuss
which Gemini free are you talking about? Gemini from aistudio.google.com ?

my guess would be asymmetrical information (most non tech people don't know it exist) and the explicit statement that google will use any prompt shared there, in its free tier, for training and improving their own models
eevmanu
·7 mesi fa·discuss
I find the premise reasonable, though I do have an observation.

The current AI hype may have placed us in a filter bubble or echo chamber, shaping our conclusions. These highly specialized algorithms can nudge or reward us for thinking in specific ways.

Regarding programming languages, there is immense value in understanding internal primitives.

As example, consider concurrency primitives. Different languages provide different levels of abstraction: high-level library support in Python, the event loop structure in JavaScript, compiler-level implementations in Rust and C++, runtime-intrinsic mechanisms in Go and Java, and virtual machine intrinsics, such as Erlang.

By viewing languages through this lens, you recognize that each implements these primitives differently, allowing you to choose the most effective tool for the job.

If your goal is to assess the short-term economic value of a technology, your logic is understandable. However, learning new languages and tools remains worthwhile. When AI agents begin invoking these tools on the fly, you may not know if a specific choice is the most effective one. Without this knowledge, you will have some gaps to challenge the AI's decision.

In the long run, making the effort to master these concepts yields far greater value as a software engineer. It enables you to understand the rationale behind applying a precise tool to a precise task.

There are valid arguments supporting various perspectives on this. However, while any approach can be useful, this discussion highlights the need for wisdom: the awareness of one's own biases. As I noted earlier, filter bubbles can distort judgment. Continuously questioning your conclusions helps ensure you move toward the best outcomes. I hope you find this recommendation useful.
eevmanu
·7 mesi fa·discuss
what about any post from old.reddit.com ? is it accessible ?
eevmanu
·8 mesi fa·discuss
In case anyone wants to confirm if this link is official, it is.

https://goo.gle/enable-preview-features

-> https://github.com/google-gemini/gemini-cli/blob/release/v0....

--> https://goo.gle/geminicli-waitlist-signup

---> https://docs.google.com/forms/d/e/1FAIpQLScQBMmnXxIYDnZhPtTP...
eevmanu
·8 mesi fa·discuss
is this local-first (w/o network call)?