HNHacker News
TopNewBestAskShowJobs

baalimago

1,295 karma · joined July 16, 2017

learning...

https://github.com/baalimago https://lorentz.app

submissionscomments
baalimago··on DeepSeek-V4-Flash Update
Very promising. So it will both keep the speed and reduced price, yet exceed performance of the quite sufficient deepseek-v4-pro?

Should be extending the lead in intelligence/cost index, as deepseek-v4-flash already were the most price efficient model, which now becomes even better. Although, in the deepseek APIs, the cost is leaking all information about codebases to China.

baalimago··on UEFA and its national associations will not participate in FIFA competitions
It seems like Europe, yet again, have to stand up for ethics, morality and try to deflect corruption. It's so tiresome.
baalimago··on Advancing the price-performance frontier with GPT‑5.6
No we kept prompts as is, just swapped model. The prompt is already quite optimized for the task. How would updating it possibly make a more intelligent model spend less tokens than a less intelligent model? Care to elaborate?
baalimago··on Advancing the price-performance frontier with GPT‑5.6
We swapped an internal system from gpt-5-mini to gpt-5.6-luna and saw no benefit but 4x cost. Sufficed to say: we swapped back to gpt-5-mini.
baalimago··on The coolest use for the Vision Pro
Or use a drone. I'm just being clever.
baalimago··on The coolest use for the Vision Pro
How's the upstairs looking?
baalimago··on Nobody knows what a used GPU cluster is worth
Run AI locally instead of in cloud
baalimago··on Nobody knows what a used GPU cluster is worth
Is this the CDOs of the AI-bubble?
baalimago··on OpenAI says its AI went rogue and launched 'unprecedented' cyber-attack
"Look, guys, our model is dangerous too..!!"
baalimago··on Advertise in ChatGPT
I don't see how this can be any worse than normal ads. Much like how Truman finds his wife Meryl odd with her strange injections of products, the ads will become obnoxious and opaque as well. Another example is how YouTubers often embed their ads into the videos themselves, which routinely gets skipped by users as second nature (check the "most played section" on timeline to see bumps where ads end).

If I ask for deep research on which running shoes to get, it's the comparison of products I'm after. Why is A better than B, etc? I don't see how adds could come into that scope either, unless ChatGPT somehow is lying (which would make me swap to a better non-hallucinating model).

If I ask "who is the most influential Swedish poet" and ChatGPT starts blabbering about some book-company who sells books from that author: okay, maybe I do want to buy it.

So: what's the problem?

baalimago··on Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
Not good enough for high-end, not cheap enough to be for low-end. Next!
baalimago··on Show HN: Sx 2.0 – Share AI skills with your team through a Dropbox folder
What's wrong with git?
baalimago··on Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k
Selfplugging a bit here: clai[0] sends only exacly what you want it to send.

[0]: https://github.com/baalimago/clai

baalimago··on Hy3
Worse than GLM-5.2, much more expensive than deepseek v4. Tough bargain.
baalimago··on EU Parliament greenlights Chat Control 1.0
As if I wasn't apathetic enough about democracy. The only voting bill worth caring about is the one in the wallet. Then you can be the source of the corruption, instead of the sink. Every man for themselves.
baalimago··on Claude Sonnet 5
Not looking great for an upcoming IPO
baalimago··on Herdr: Agent multiplexer that lives in your terminal
Depends on the harness I imagine. If there's some sort of "post run" hook I'm sure it can be added there. Or, if the harness is open source, a PR to add it would work too.
baalimago··on Herdr: Agent multiplexer that lives in your terminal
You use an agentic harness which terminates with bell character[0]. This lights up tmux, clearly indicating that it's done.

[0]: https://en.wikipedia.org/wiki/Bell_character

baalimago··on Open source AI must win
> If intelligence becomes something people can only rent from a few closed institutions

Just your your natural born intelligence..? It's worked for the past 10k+ years, I'm sure it will work for some time longer

baalimago··on Statement on US government directive to suspend access to Fable 5 and Mythos 5
International customers might not be so keen on buying Anthropic, xAi or OpenAI products if they can be disrupted by the US government like this. The market within USA is surely not large enough to live up to the financial promises that keep this AI bubble growing.

Does the White House want the AI bubble to pop..? Incredibly dumb move.

baalimago··on Why AI hasn't replaced software engineers, and won't
That's the point: they would have. Now they spend money on tokens instead (and a equity to me who set up their environment)
baalimago··on Why AI hasn't replaced software engineers, and won't
It most certainly will replace software engineers. What's missing is, as the article suggests, the "Delivery" bit. But that's not the realm of software engineers, that's the realm of DevOps/SRE/Cloud engineers.

I work as a cloud engineer and have been contacted by multiple non-engineering friends who have now been able to create their pet projects from scratch in different languages and have it running locally, as webapps and native apps. So what they are missing is a platform to easily deploy and maintain their projects, much like a "normal" developer would. Right now it's quite tedious to set up this scaffolding, but it's absolutely possible with AGENTS.md, skills and rigid hollistic tests. Once done, non-technical people can continue developing independently without hiring any software engineers by simply telling claude/codex what they want. Claude/codex will then be able to make judgement calls based on the preset architecture, which will guide the non-technical user.

So in my anecdotal case, AI has already replaced several software engineers. Once scaffolding like this is productized, I suspect that greenfield projects can be managed entirely from a product standpoint using agentic coders + platform engineering. And that is today. Imagine in 5 years.

baalimago··on πFS
This got me thinking about the "simulation theory":

If our universe is simulated, it must be possible to snapshot the entire state for one iteration (however time now is quantized, open question). "... From here, it is a small leap to see that if π contains all possible files, why are we wasting exabytes of space storing those files, when we could just look them up in π!" (from pifs, above)

This means that not only does a singular snapshot of our universe exists in pi, but every single one does

The information for our entire universe's simulation is stored in pi (and every other number like it)

baalimago··on Claude Fable 5
I can't justify a pricetag like that when deepseek v4 pro is $0.003625/1M for cache hit, $0.435 for cache miss and $0.87 /1M tokens for output.

For the token cost of explaining some task to Fable, deepseek v4 pro is able to solve the same task many times over.

baalimago··on Zig by Example
I'm missing the concurrency model.
baalimago··on Gemma 4 12B: A unified, encoder-free multimodal model
I don't understand why Google does this. If I can run this locally, why would I need a subscription or use any inference provider, including Google..?

Scorched earth tactics to make anthropic and openai IPO fail?

baalimago··on Incident with Pull Requests, Issues, Git Operations and API Requests
Perhaps best to simply declare indefinite-mortem
baalimago··on All of human cooking compressed into 2 megabytes
Great, so now chefs are being replaced too..!
baalimago··on Incident with Actions and Pages
Hey at least Copilot AI Model Providers have 100% uptime, so there's that
baalimago··on I spent 50 hours drawing a line graph
One day I wish to be in a stable enough place that it's possible for me to pursue activities like these. It's a true privilege
← PreviousPage 3 of 13Next →