jryio

6 ms·

31.
by jryio 5mo ago
If anyone was wondering ... it's racist Unsurprisingly the texts written up until that time were dominated by such individuals which is tragic for LLM training if you think about it. The voiceless groups or fringe opinions which we tak
32.
by jryio 5mo ago
> OpenAI has contracted to purchase an incremental $250B of Azure services, and Microsoft will no longer have a right of first refusal to be OpenAI’s compute provider. Azure is effectively OpenAI's personal compute cluster at this s
33.
by jryio 6mo ago
I find it strange that you've anthropomorphized Claude but not ChatGPT seemingly based on one having a human name and the other not
34.
Show HN: Obscura – V8-powered headless browser for scraping and AI agents(github.com)
15 points by jryio 6mo ago|0 comments
35.
by jryio 6mo ago
Exactly - cooperation is not incentivized properly
36.
by jryio 6mo ago
Just another disposable piece of software maintained by a single person that does 80% of what other apps do but worse. Max lifespan 2 years
37.
by jryio 6mo ago
This feels like an unethical release of a model. They've opened a can of worms without investing in defense first. Anthropic announced their capabilities in advanced, issued a private release, then put up $100M in credits to Fortune 50
38.
by jryio 6mo ago
Model performance at inference in a data center v.s. stripping thinking tokens are effectively the same. Sure they didn't change the GPUs their running, or the quantization, but if valuable information is removed leading to models perf
39.
by jryio 6mo ago
Their 'Preparedness Framework'[1] is 20 pages and looks ChatGPT generated, I don't feel prepared reading it. https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbdde...
40.
by jryio 6mo ago
1. They changed the default in March from high to medium, however Claude Code still showed high (took 1 month 3 days to notice and remediate) 2. Old sessions had the thinking tokens stripped, resuming the session made Claude stupid (took 15
41.
GitHub Ace - an Experimental Collaboraite Agentic Coding Environment(maggieappleton.com)
6 points by jryio 6mo ago|0 comments
42.
by jryio 6mo ago
Notion did it first and arguably better[1]. Shared agents benefit from shared context. The hardest part is ensuring that shared context is maintained and it converges on a representation of reality and the people in the company. [1] https:
43.
by jryio 6mo ago
Software engineering is certainly not engineering. Even at the highest levels. Real engineering have infinitely more complex interactions in the physical world than symbolic institutions for machines.
44.
by jryio 6mo ago
No one was discussing 'taste' until pg's article. After a decade with tech people I can confidently say that most of them have zero taste because they have little to no exposure to the world outside of their bubble. It's
45.
by jryio 6mo ago
The majority of vulnerabilities are in newly committed lines of code. This has been shown again and again [1] [2] From a marketing standpoint Anthropic is showing that they're able to direct 'compute' to find vulnerabilities
46.
by jryio 6mo ago
Let's fast forward the clock. Does software security converge on a world with fewer vulnerabilities or more? I'm not sure it converges equally in all places. My understanding is that the pre-AI distribution of software quality (an
47.
Is All Software Converging(jry.io)
3 points by jryio 6mo ago|0 comments
48.
by jryio 6mo ago
Nicely done! I didn't catch if you were cited.
49.
Embarrassingly Simple Self-Distillation Improves Code Generation(arxiv.org)
4 points by jryio 6mo ago|3 comments
50.
by jryio 6mo ago
Here's an interesting discussion from Section 8 - Dormant Wallets: If a nation state develops a sufficiently powerful quantum computer. Seizure of the Satoshi-era bitcoin wallets without post quantum protections would fund either rogue
51.
by jryio 6mo ago
The godoc format enforces that the comment start with the name of the identifier and be a complete sentence(s) describing what that identifier does. Predates LLMs
52.
by jryio 6mo ago
It's exhausting to make this comment every time... but here we go. Key revocation is table stakes for secure messaging. I need a trusted way to relay that my contact's key has been revoked and I should stop trusting it. Neither P2
53.
by jryio 7mo ago
If you're reading this, we're sorry John!
54.
by jryio 7mo ago
Someone's human and likes typos. Might be the last signal of humanity online if you think about it .
55.
by jryio 7mo ago
I imagine you live your life contextually, whereby your daily experiences are felt against the backdrop of the immediate events you, then your community, and eventually the world at large. If the rest of the world was involved, why not brin
56.
by jryio 7mo ago
I am commenting on the entire app experience on iOS not a single web app they released today (which unfortunately is what can be linked on HN). Read the other comments and you'll see the same, download the iOS app and use that as your
57.
by jryio 7mo ago
Flighty is a good representation of what craft - compounded over time - gives you. Everything from on design, to features, to data integrations. It's everything that vibe coding and agents don't get you. I appreciate their craft.
58.
by jryio 7mo ago
When Apple vertically integrates it works for them. All the way from the cloud to the OS to the hardware. Pretty sure this will beat out tools like JAMF on user privacy alone by running trusted MDM adjacent tools in kernel space. Yes sure y
59.
by jryio 7mo ago
Nice upgrade. userpsace HTTP proxies are a good start and should make unlikely that a secret gets into the context window due to a high permission read. There are a few missing pieces in the agent security world in general 1. Full secret-me
60.
by jryio 7mo ago
Thanks for the great feedback. There's a link to find some time when you're passing through. Otherwise shoot me an email! Cheers.

More ›