brendanashworth

This link gives 404 Page not found, perhaps https://amplitude.com/blog/amplitude-and-statsig-partnership is a better URL [1]

[1] https://news.ycombinator.com/item?id=48028790

Out of the box, Opus/GPT-5.4 uses checks like `tsc --noEmit` or `python -m py_compile ...`, so compile checks on a Swift/Objective-C app probably get you pretty far. Also just setting up a Dockerfile maybe with swift:5.10-focal would give the agent the right tools to verify its own work.

Sandboxing isn't about having the perfect devex environment, most dev sandboxes (looking at codex cloud) don't actually have full verification available. Sometimes Github actions CICD is all you can get!

Is there a reason people don't use SHAP [1] to interpret language models more often? The in-context attribution of outputs seems very similar.

[1] https://shap.readthedocs.io/en/latest/

If choosing a Bayesian approach in a clinical trial can reduce the number of recruited subjects, I would imagine the pharma industry is strongly incentivized to adopt it.

Would a wearable model like this gain in predictive power by adding FHIR/EHR inputs? CoMET [1] has great AUC even without daily/lifestyle metrics like steps, awakenings at night, or vo2max estimates (which are present in say apple health). It would be interesting to know how much you would gain from having a holistic model that sees both inputs.

[1] https://arxiv.org/pdf/2508.12104v1

How does this compare to obstore? [1]

[1] https://developmentseed.org/obstore/latest/

Cursor [1] has excellent codebase RAG, and great recall of third party documentation.

Use "instruction @web @Flask" to search the internet and Flask docs before answering, for example.

[1] https://www.cursor.com/

Property boundaries aren't delimited by GPS coordinates, but rather a surveyor's instructions on how to locate the boundary from a set of known landmarks. In most US states, the Public Land Survey System [1] gives "meridians" and "baselines" for each region serving as the regional center. Radiating out, every 6 miles, is a new "township," with its corners serving as the local landmark for property boundaries.

[1] https://en.wikipedia.org/wiki/Public_Land_Survey_System

I am confused by the benchmarks you provided.

(1) The "baseline" comparison was to PyPDF + Naive RAG. For the LlamaParse evaluation, you appear to have used a different RAG pipeline, called "recursive retrieval." Why not use the same pipeline to demonstrate the improvement from LlamaParse? Can you share the code to your evaluation for LlamaParse?

(2) I ran the benchmark for the PyPDF + Naive RAG solution, directly copying the code on the linked LlamaIndex repo [1]

I got very different numbers: mean_correctness_score 3.941 mean_relevancy_score 0.826 mean_faithfulness_score 0.980

You reported: mean_correctness_score 3.874 mean_relevancy_score 0.844 mean_faithfulness_score 0.667

Notably, the faithfulness score I measured for the baseline solution was actually higher than that reported for your proprietary LlamaParse based solution.

[1] https://github.com/run-llama/llama-hub/tree/main/llama_hub/l...

This could be part of a great contribution to the Natural Earth project [1], which is one of the major sources of cartographic raster data for online web maps!

[1] https://www.naturalearthdata.com/

Reasonable base layers for maps are table stakes for geospatial apps, so it's great to see Protomaps getting some attention. Requiring an expensive MapBox subscription (pay per tile render) for popular apps just meant that it became prohibitively expensive to run side projects with maps.

If you're interested in cartography, Facebook created the Daylight Map Distribution [1] project. They analyzed satellite imagery, outlined buildings and roads, and published them for map makers. It significantly improves coverage in rural areas and non-Western countries.

[1] https://daylightmap.org

Working within the existing system, though, is typically the fastest way to address the grievance. AI is probably the most scalable approach to competent counsel, far easier (given ChatGPT) than training new lawyers and convincing them to become public defenders.

I think it's good to recognize that both enterprises and individuals want to use and contribute to large projects like OpenStreetMap.

Overture's sponsors seem pretty focused around big company needs, like coordinating efforts between mapping orgs and standardizing metadata. They're each pledging $3M + 20 FTEs per year to accomplish it [1].

But I would argue it's more productive to coordinate enterprise+individual contributions through a central organization: the OSM foundation.

When io.js forked from Node.js in 2014, it was more of a call for attention than a desire to split the ecosystem in two. They merged the next year. (I joined the core dev team after the split.)

Maybe instead of asking what OSM will become after Overture, we should ask, why can't enterprises accomplish their goals within the framework of the OSM Foundation?

Disclaimer: my startup has a product built on OSM [2].

[1] https://overturemaps.org/become-a-member/

[2] https://buntinglabs.com/solutions/openstreetmap-extracts

The best article I've read about Katerra is still from Construction Physics, it goes into much more technical detail. [1]

[1] https://constructionphysics.substack.com/p/another-day-in-ka...

Definitely agree with what you're saying here! Maybe I wouldn't say its for beginner programmers - many might immediately be scared by the idea of a bitwise operator or raw memory access - it should definitely be somewhere in the ladder for learning new languages.

Some people simply don't understand that C is actually everywhere around them. They don't understand that either the language they use is built in C, or the language THAT was built on was built in / on C, etc etc. Simply the importance of the language in history may just be a powerful point on its own. If someone knows quite a few languages and doesn't yet know C, I kind of don't understand why - they do not yet understand the importance of lower level operations, like the functioning of a garbage collector or why their pre-mature optimizations shouldn't even be attempted.

Great article, over all - I really liked it.