jakozaur

6 ms·

31.
by jakozaur 8mo ago
Funny coincidence, I'm working on a benchmark showcasing AI capabilities in binary analysis. Actually, AI has huge potential for superhuman capabilities in reverse engineering. This is an extremely tedious job with low productivity. Cu
32.
by jakozaur 8mo ago
Great idea. Currently, people have to rely on client-side spans in OpenTelemetry. However, it would be awesome if we could get spans for slow SQL queries, along with explanations.
33.
by jakozaur 8mo ago
In this benchmark, micro-services are really small, ~300 lines, and sometimes just two of them. More realistic tasks (large codebases, more microservices) would have a lower success rate.
34.
OTelBench: Can AI instrument OpenTelemetry?(quesma.com)
5 points by jakozaur 9mo ago|1 comments
35.
by jakozaur 9mo ago
See x thread for rationale: https://x.com/mitchellh/status/2014433315261124760?s=46&t=FU... “ Ultimately, I want to see full session transcripts, but we don't have enough tool support for that broadly.” I
36.
by jakozaur 10mo ago
I wish Neal would do behind the scenes, how he built this art. I wonder whether LLM assistants like Claude Code make such an interactive show more feasible. He previously did a game "Infinite Craft" which leveraged Llama models. H
37.
by jakozaur 10mo ago
Not sure if I get this: WASM lets you use any language in the browser, though it still works way better with languages without GC, such as Rust or a transpiling C engine. Java is unlikely to be the best choice. In the era of LLM assistants
38.
by jakozaur 10mo ago
LLVM IR is quite fun to play with from many programming languages. The Java example is rather educational, but there are several practical example,s such as in Go Lang: https://github.com/llir/llvm
39.
by jakozaur 10mo ago
The effect of climate change may be highly uneven. Some regions will be fine with adaptation, while other places will hardly sustain cities.
40.
by jakozaur 10mo ago
It is even more true with startups and business. Super rushed is bad, but doing for too long decreases quality.
41.
by jakozaur 10mo ago
Looks like Arc, would love to migrate out of it after migration, but always worry about maintenance. Creating a browser is "easy", keeping it up to date is a lot of work, and many open-source browsers look semi-abandoned to me.
42.
by jakozaur 10mo ago
Is it just me, or do LLM code assistants do catastrophically silly things (drop a DB, delete files, wipe a disk, etc.) far more often than humans? It looks like the training data has plenty of those examples, but the models don’t have enoug
43.
Nano Banana Pro: raw intelligence with tool use(quesma.com)
4 points by jakozaur 11mo ago|0 comments
44.
by jakozaur 11mo ago
FLUX.1 Pro Kontext was one of the best artistic model, still great at instruction following comparing to MidJourney V7. See my third comparison in Nano Banana blog post: https://quesma.com/blog/nano-banana-pro-intellig
45.
A postmortem on our $2.5M database gateway: lessons from pilot purgatory(quesma.com)
6 points by jakozaur 11mo ago|0 comments
46.
by jakozaur 11mo ago
Source: https://seekingalpha.com/article/4830274-salesforce-inc-crm-...
47.
by jakozaur 11mo ago
I talked to some enterprises and saw similar patterns: 1. Agentic AI systems are hard to measure and evaluate methodologically. 2. Quote from Salesforce analyst day: "it's been so easy to build a killer demo, but why has it been s
48.
The security paradox of local LLMs(quesma.com)
160 points by jakozaur 1y ago|87 comments
49.
Local LLMs are worse for security(quesma.com)
1 points by jakozaur 1y ago|0 comments
50.
by jakozaur 1y ago
(Tech) Debt seems to be a frequent cause of (Tech) outages.
51.
by jakozaur 1y ago
Just use `claude update` if you already have it. Unfortunately, they removed Plan mode, when I could use Opus for planning and Sonnect for coding. Though I will see how this pans out.
52.
by jakozaur 1y ago
You can be a a CEO of $79Bln market cap company (CloudFlare) and still post on Hacker News. Funny thing, his initial Hacker News submissions from 2010 and 2011 received fewer than 10 upvotes, but somehow he wasn't discouraged.
53.
CompileBench: Can AI Compile 22-year-old Code?(quesma.com)
148 points by jakozaur 1y ago|65 comments
54.
Winners of OpenAI GPT-OSS-20B Red‑Teaming Challenge(kaggle.com)
1 points by jakozaur 1y ago|0 comments
55.
by jakozaur 1y ago
Does it apply to people who planned to start on Oct 1, 2025? With the current system, you must apply in April if you succeed in the lottery, and then you can start in a few months in October, once per year. Looks very uncomfortable for thos
56.
by jakozaur 1y ago
A Ponzi scheme is extreme, where the underlying asset is worthless. Databricks is a fast-growing company with ~$4B in annualised revenue and huge potential. Many rounds got some portion of the round for liquidity. Similarly, markup strategi
57.
by jakozaur 1y ago
It doesn't look like a typical round for raising capital for investments. Instead: 1. Liquidity: Early investors could sell to late-stage investors, since they are not IPO. Their previous round looked like that. 2. Markup: The previous
58.
by jakozaur 1y ago
The coding seems to be one of the strongest use cases for LLMs. Though currently they are eating too many tokens to be profitable. So perhaps these local models could offload some tasks to local computers. E.g. Hybrid architecture. Local mo
59.
by jakozaur 1y ago
This Twitter account is an influencer with 200,000+ followers, known for its hot takes. Though the risk of GenAI is real, it looks to me there is a fair amount of chance that this story is staged and amplified for social media drama purpose
60.
by jakozaur 1y ago
This feels similar to TurboBuffer, which is also built on top of S3 storage. TurboBuffer has been a leader in this space, powering vector search for major companies like Cursor, Linear, and Notion. It seems AWS is leveraging its strong S3 b

More ›