nicola_alessi

7 ms·

1.
Ask HN: Opus 4.7 – is anyone measuring the real token cost on agentic tasks?
1 points by nicola_alessi 6mo ago|0 comments
2.
by nicola_alessi 7mo ago
Fair point, appreciate the callout. I'll dial it back.
3.
by nicola_alessi 7mo ago
Interesting framing — hadn't thought about it from the inference routing angle but it maps well to what the data shows. On latency variance: yes, significantly. Cost standard deviation across runs dropped 6-24x depending on task type.
4.
More tokens, less cost: why optimizing for token count is wrong
1 points by nicola_alessi 7mo ago|9 comments
5.
Show HN: Focused input cuts LLM output tokens by 63% bench on CC with FastAPI
2 points by nicola_alessi 7mo ago|0 comments
6.
by nicola_alessi 7mo ago
vexp works regardless of how you organize files. The graph is at the symbol level (functions, classes, types) not the file level, so whether you have one function per file or 50, it resolves the same dependency chain and serves the same cap
7.
Show HN: 58% cost by replacing file reads with a dependency graph on AI Coding
4 points by nicola_alessi 7mo ago|2 comments
8.
by nicola_alessi 7mo ago
vexp is an MCP server that gives AI coding agents structural awareness of your codebase. Instead of letting the agent read entire files, it pre-computes a dependency graph with tree-sitter and serves only the relevant code nodes — full sour
9.
by nicola_alessi 7mo ago
Gating completion on the observation write is smart — you're turning the model's task-completion drive against itself. Have you run into the problem where forced observations degrade to "completed task successfully, no issues
10.
Ask HN: Why do AI coding agents refuse to save their own observations?
2 points by nicola_alessi 7mo ago|1 comments
11.
Show HN: Vexp – Your AI coding agent forgets everything. Mine doesn't(vexp.dev)
1 points by nicola_alessi 8mo ago|0 comments
12.
Show HN: Vexp – Local-first context engine for AI coding agents
1 points by nicola_alessi 8mo ago|0 comments
13.
Show HN: Vexp – graph-RAG context engine, 65-70% fewer tokens for AI agents
3 points by nicola_alessi 8mo ago|0 comments
14.
by nicola_alessi 10mo ago
You are 100% correct, and this is the central limitation. An LLM like ChatGPT, trained on general web text, is a terrible movie recommendation engine for exactly the reasons you state. Its knowledge is broad but shallow, skewed toward popul
15.
by nicola_alessi 10mo ago
This is a fantastic point, and you've hit on something fundamental that's been lost in the shift to on-demand: the joy of discovery through serendipity and low commitment. You're describing the exploration/exploitation t
16.
by nicola_alessi 10mo ago
Your point is excellent and cuts to the core of what we're trying to explore. You're right, ‘mood' can be a fuzzy, high-friction starting point. The hypothesis behind the prompt isn't that everyone consciously identifies
17.
Is Entertainment Discovery Fundamentally Broken?
2 points by nicola_alessi 10mo ago|9 comments
18.
An experiment in mood-based movie discovery: Lumigo.tv(lumigo.tv)
1 points by nicola_alessi 11mo ago|1 comments
19.
by nicola_alessi 11mo ago
I’ve been testing a platform called Lumigo.tv, and it made me rethink how recommendation systems could work if they started from human emotion instead of metadata. The core idea is simple: instead of browsing genres or relying on collaborat
20.
A Scalable Standard for Clean ECommerce Data in LLMs (Fork of Llms.txt)
2 points by nicola_alessi 1y ago|0 comments
21.
by nicola_alessi 1y ago
You're absolutely right that existing standards like sitemap.xml + Accept headers could work in theory, but here's why we built this for eCommerce specifically: The HTML-to-Markdown Problem Even with Accept: text/markdown, mo
22.
Site-Llms.xml – An XML Sitemap Standard for AI-Friendly ECommerce Data(github.com)
1 points by nicola_alessi 1y ago|3 comments
23.
Ask HN: Why hasn't anyone fixed product search yet?
2 points by nicola_alessi 2y ago|0 comments
24.
by nicola_alessi 2y ago
A year ago, I bought a ‘top-rated’ DSLR camera based on Google’s first-page results. It turned out to be a refurbished model sold as new. That frustration led to Lumigo ( https://lumigo.ai ), an AI search engine that: Bans all ads
25.
by nicola_alessi 2y ago
I'm frustrated with: First-page Google results being 80% ads Amazon reviews gaming the system No transparency in recommendations So I built Lumigo Lumigo that: Only shows merit-based results Cites all sources (blogs/forums/re
26.
by nicola_alessi 2y ago
we want to move in this direction: - offer subscriptions for merchants and earn percentages on sales by implementing a direct checkout (but the fact that lumigo forms a partnership with brand X is not an indicator that the algorithm will be
27.
by nicola_alessi 2y ago
Hi HN, I’ve been working on Lumigo, an AI-powered search engine designed to fix the mess of online product search. No ads, no fake reviews—just honest, transparent recommendations based on real data. We’re launching soon on Product Hunt, an
28.
by nicola_alessi 2y ago
yes, of course, I recognise that it is quite tricky to distinguish reliable products from those that can be defined as ‘junk’... that's precisely why I thought of giving each search a large amount of data referring to the product you a
29.
by nicola_alessi 2y ago
it depends on the sources that are used and above all on how much data is imagined before making an average ranking of the products... so in that case it is very useful rather than reading N sources and then deciding what to focus on
30.
Is Product Search Broken? Why Are We Still Stuck with Ads and Fake Reviews?
27 points by nicola_alessi 2y ago|35 comments

More ›