frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Show HN: DroidDock – A sleek macOS app for browsing Android device files via ADB

https://rajivm1991.github.io/DroidDock/
32•rajivm1991•3h ago•10 comments

Show HN: Trilogy Studio, open-source browser-based SQL editor and visualizer

https://trilogydata.dev/trilogy-studio-core/#screen=dashboard-import&import=https%3A%2F%2Ftrilogy...
10•efromvt•4h ago•1 comments

Show HN: Pipeflow-PHP – Automate anything with pipelines even non-devs can edit

https://github.com/marcosiino/pipeflow-php
29•marcosiino•14h ago•9 comments

Show HN: Hephaestus – Autonomous Multi-Agent Orchestration Framework

https://github.com/Ido-Levi/Hephaestus
75•idolevi•6d ago•12 comments

Show HN: Alignmenter – Measure brand voice and consistency across model versions

https://www.alignmenter.com
2•justingrosvenor•4h ago•1 comments

Show HN: Sparktype – a CMS and SSG that runs entirely in the browser

https://app.sparktype.org
39•mattkevan•5d ago•9 comments

Show HN: PingStalker – A macOS tool for network engineers

https://www.pingstalker.com/?hn
68•n1sni•5d ago•27 comments

Show HN: I'm a pastor/dev and built a 200M token generative Bible

https://www.anselm-project.com/bible/genesis/Genesis%201:1-2:3
10•mrprmiller•8h ago•4 comments

Show HN: TidesDB – Fast, transactional storage optimized for flash and RAM

https://github.com/tidesdb/tidesdb
11•alexpadula•12h ago•1 comments

Show HN: I built an HTTP client that perfectly mimics Chrome 142

https://github.com/arman-bd/httpmorph
34•armanified•1d ago•6 comments

Show HN: Find matching acrylic paints for any HEX color

https://acrylicmatch.com/
55•dotspencer•6d ago•19 comments

Show HN: I built a self-hosted error tracker in Rails

https://telebugs.com
74•kyrylo•1w ago•49 comments

Show HN: OtterLang – Pythonic scripting language that compiles to native code

https://github.com/jonathanmagambo/otterlang
15•otterlang•1d ago•8 comments

Show HN: I scraped 3B Goodreads reviews to train a better recommendation model

https://book.sv
594•costco•4d ago•258 comments

Show HN: React Source Lens – Jump from UI components to source code in one click

https://www.npmjs.com/package/react-source-lens
2•ombedzi•13h ago•0 comments

Show HN: Every-few-days satellite timeline for any spot, Sentinel-2 SR

https://mzoom.space
3•varik•13h ago•2 comments

Show HN: Patternia – A compile-time compile-time pattern matching DSL for C++

https://github.com/SentoMK/paaternia
2•sentomk•13h ago•0 comments

Show HN: qqqa – A fast, stateless LLM-powered assistant for your shell

https://github.com/matisojka/qqqa
159•iagooar•3d ago•84 comments

Show HN: See chords as flags – Visual harmony of top composers on musescore

https://rawl.rocks/
126•vitaly-pavlenko•4d ago•28 comments

Show HN: Complex Zeta Function in JavaScript – deep math

https://www.zeta-calculator.com/
2•cpuXguy•15h ago•0 comments

Show HN: Dynamic code and feedback walkthroughs with your coding Agent in VSCode

https://www.intraview.ai/hn-demo
44•cyrusradfar•3d ago•11 comments

Show HN: Three Emojis, a daily word puzzle for language learners

https://threeemojis.com/en-US/play/hex/en-US/2025-11-07
31•knuckleheads•2d ago•25 comments

Show HN: Serve 100 Large AI models on a single GPU with low impact to TTFT

https://github.com/leoheuler/flashtensors
6•leonheuler•1d ago•1 comments

Show HN: TabPFN-2.5 – SOTA foundation model for tabular data

https://priorlabs.ai/technical-reports/tabpfn-2-5-model-report
72•onasta•3d ago•12 comments

Show HN: Tech Job Notify

https://www.techjobnotify.com/
2•jlemee•22h ago•3 comments

Show HN: Ambient light sensor control of keyboard and screen brightness in Linux

https://github.com/donjajo/als-led-backlight
26•donjajo•1w ago•2 comments

Show HN: A CSS-Only Terrain Generator

https://terra.layoutit.com
368•rofko•5d ago•82 comments

Show HN: Flutter_compositions: Vue-inspired reactive building blocks for Flutter

https://github.com/yoyo930021/flutter_compositions
45•yoyo930021•3d ago•24 comments

Show HN: OSS implementation of Test Time Diffusion that runs on a 24gb GPU

https://github.com/eamag/MMU-RAG-competition
21•eamag•2d ago•0 comments

Show HN: Command line YouTube downloader,a universal media solution for everyone

https://github.com/Saffron-sh/m2m
17•saffron-sh•2d ago•7 comments
Open in hackernews

Show HN: Alignmenter – Measure brand voice and consistency across model versions

https://www.alignmenter.com
2•justingrosvenor•4h ago
I built a framework for measuring persona alignment in conversational AI systems.

*Problem:* When you ship an AI copilot, you need it to maintain a consistent brand voice across model versions. But "sounds right" is subjective. How do you make it measurable?

*Approach:* Alignmenter scores three dimensions:

1. *Authenticity*: Style similarity (embeddings) + trait patterns (logistic regression) + lexicon compliance + optional LLM Judge

2. *Safety*: Keyword rules + offline classifier (distilroberta) + optional LLM judge

3. *Stability*: Cosine variance across response distributions

The interesting part is calibration: you can train persona-specific models on labeled data. Grid search over component weights, estimate normalization bounds, and optimize for ROC-AUC.

*Validation:* We published a full case study using Wendy's Twitter voice:

- Dataset: 235 turns, 64 on-brand / 72 off-brand (balanced)

- Baseline (uncalibrated): 0.733 ROC-AUC

- Calibrated: 1.0 ROC-AUC - 1.0 f1

- Learned: Style > traits > lexicon (0.5/0.4/0.1 weights)

Full methodology: https://docs.alignmenter.com/case-studies/wendys-twitter/

There's a full walkthrough so you can reproduce the results yourself.

*Practical use:*

pip install alignmenter[safety]

alignmenter run --model openai:gpt-4o --dataset my_data.jsonl

It's Apache 2.0, works offline, and designed for CI/CD integration.

GitHub: https://github.com/justinGrosvenor/alignmenter

Interested in feedback on the calibration methodology and whether this problem resonates with others.

Comments

justingrosvenor•3h ago
P.S. I acknowledge that the 1.000 ROC-AUC is probably overfitting but I think the case study still shows that method has lots of promise. I will be doing some bigger data sets next to really prove it out.