H
HN HRCB stories | rights | sources | trends | system | about
System Pipeline operations and system health. connecting...
Pipeline
320
Done
999
Pending
0
Evaluating
2
Failed
1668
Skipped
320.0 evals/day 1 days active est. 4 days to clear backlog
Multi-Model Raters
Model Avg Score Conf Done Queued Failed Coverage
claude-haiku-4-5-20251001 0.187 0.20 271 0 30
85%
deepseek/deepseek-v3.2-20251201 0.146 0.12 335 80 4
105%
meta-llama/llama-3.3-70b-instruct:free --- --- 0 348 0
0%
@cf/meta/llama-4-scout-17b-16e-instruct 0.026 0.59 246 0 32
77%
nvidia/nemotron-3-nano-30b-a3b:free --- --- 0 0 97
0%
arcee-ai/trinity-large-preview:free 0.031 0.16 34 0 0
11%
Evaluation Models 5 models in history
Model Status Provider Evals Avg HRCB Min Max
claude-haiku-4-5-20251001 on anth 320 +0.18 -0.65 +0.86
deepseek/deepseek-v3.2-20251201 on OR 336 +0.15 -0.85 +0.81
arcee-ai/trinity-large-preview:free off OR 35 +0.03 -0.30 +0.35
nvidia/nemotron-3-nano-30b-a3b:free on OR 2 0.00 0.00 0.00
stepfun/step-3.5-flash:free off OR --- --- --- ---
qwen/qwen3-next-80b-a3b-instruct:free off OR --- --- --- ---
meta-llama/llama-3.3-70b-instruct:free on OR --- --- --- ---
mistralai/mistral-small-3.1-24b-instruct:free off OR --- --- --- ---
nousresearch/hermes-3-llama-3.1-405b:free off OR --- --- --- ---
@cf/meta/llama-3.3-70b-instruct-fp8-fast off CF --- --- --- ---
@cf/meta/llama-4-scout-17b-16e-instruct on CF 253 +0.03 -0.80 +0.80
Pipeline Events 282892 last 24h · 282923 last 7d · 271872 errors
Errors/warnings (7d)
API Headroom claude-haiku-4-5-20251001 · 2026-02-26 04:51
Requests
999/1000
Input Tokens
449k
Output Tokens
89k
Cache: 69% 429s: 0
credit_exhausted: 169603 dlq: 97524 rate_limit: 6683 eval_failure: 2494 eval_retry: 2142 cron_run: 1475 eval_success: 1420 self_throttle: 706 dlq_replay: 602 rater_validation_fail: 100 rater_validation_warn: 63 coverage_crawl: 38 trigger: 32 cron_error: 22 rater_auto_disable: 12 story_dead: 3 calibration: 2 eval_skip: 2
Cycle Performance
Cycle Start Crawl Found New Evals Fail Cycle Time
2026-02-26 23:15 17.9s 712 3 0 --- current
2026-02-26 23:14 8.7s 712 0 1 --- 1m 5s
2026-02-26 23:13 12.4s 712 0 17 --- 1m 0s
2026-02-26 23:11 6.9s 712 3 20 --- 1m 6s
2026-02-26 23:11 16.4s 712 1 8 --- 0m 40s
2026-02-26 23:10 10.3s 712 1 1 --- 1m 17s
2026-02-26 23:09 10.1s 712 1 0 --- 0m 37s
2026-02-26 23:07 6.5s 712 2 20 --- 1m 26s
2026-02-26 23:06 6.6s 712 0 19 --- 1m 0s
2026-02-26 23:06 13.1s 712 1 11 --- 0m 43s
Avg: 10.8 evals/cycle · 0m 59s cycle time
Dead Letter Queue
22 pending · 602 replayed · 96905 discarded
47110393 The Work Behind the Writing: O...
47172203 Block shares soar 24% as compa...
47128645 I baked a pie every day for a ...
Methodology Versions
9d3c9ac6 · 246 evals
pre-v · 74 evals (stale)
74 evals need reprocessing
Model Drift
claude-haiku-4-5-20251001 · 902 evals · avg 0.213 ±0.262
deepseek-v3.2 · 336 evals · avg 0.148 ±0.245
llama-4-scout-wai · 253 evals · avg 0.027 ±0.199
trinity-large · 35 evals · avg 0.031 ±0.102
nemotron-nano-30b · 2 evals · avg 0.000 ±0.000
286 overlapping · mean |Δ| = 0.198
Calibration
No runs yet
15 URLs defined
Recent Events
2026-02-26 23:15 cron_run Cron: 3 new, 712 unique stories - -
2026-02-26 23:14 credit_exhausted Credit balance too low, pausing provider for 30 min - -
2026-02-26 23:14 eval_success Light evaluated: Neutral (0.00) - -
2026-02-26 23:14 cron_run Cron: 0 new, 712 unique stories - -
2026-02-26 23:14 eval_success Light evaluated: Neutral (0.00) - -
2026-02-26 23:13 eval_success Light evaluated: Neutral (0.00) - -
2026-02-26 23:13 eval_success Light evaluated: Mild negative (-0.20) - -
2026-02-26 23:13 rater_validation_fail Light parse failure for model llama-4-scout-wai: SyntaxError: Unexpected token '+', ..."itorial": +0.6, "... is not valid JSON - -
2026-02-26 23:13 eval_success Light evaluated: Neutral (0.00) - -
2026-02-26 23:13 eval_success Light evaluated: Neutral (0.00) - -
2026-02-26 23:13 eval_success Light evaluated: Moderate positive (0.50) - -
2026-02-26 23:13 eval_success Light evaluated: Moderate positive (0.40) - -
2026-02-26 23:13 eval_success Light evaluated: Mild negative (-0.20) - -
2026-02-26 23:13 rater_validation_warn Light validation warnings for model llama-4-scout-wai: 1W 0R - -
2026-02-26 23:13 eval_success Light evaluated: Moderate positive (0.60) - -
Recent Failures
What Claude Code Chooses (amplifying.ai) Error: Anthropic API error 400: {"type":"error","error":{"type":"invalid_request
New York sues Valve for enabling "illegal gambling" with loot boxes (arstechnica.com) no scores for rescoring
Queue 100 stories waiting · ~20 per batch · est. 50 min to clear
# Status Story Domain
1 pending Statement from Dario Amodei on Our Discussions with the Department of War www.anthropic.com
2 pending Metacritic statement pledges to ban outlets that use AI-generated reviews www.shacknews.com
3 pending Hillary Clinton's Opening Statement to House Oversight and Gov Reform Committee twitter.com
4 pending Hydroph0bia – a fixed SecureBoot bypass for UEFI firmware based on Insyde H2O coderush.me
5 pending Smartphone Mkt to Decline 13% in '26, Largest Drop Ever Due to Memory Shortage www.idc.com
6 pending Canadian government demands safety changes from OpenAI www.engadget.com
7 pending Show HN: Usplus.ai – Build a company of AI agents and execute work autonomously usplus.ai
8 pending Block (Square) plans to lay off nearly half its staff in embrace of AI www.morningstar.com
9 pending Show HN: Stop reviewing AI-generated code during a PR, move it in the edit cycle medium.com
10 pending Kansas invalidates drivers licenses of trans people www.theguardian.com
11 pending Show HN: Safari-CLI – Control Safari without an MCP www.npmjs.com
12 pending The Remote-Work Dream Isn't Dead, but It's Slipping Away www.wsj.com
13 pending iPhone and iPad Are First Consumer Devices Cleared for NATO Classified Data www.macrumors.com
14 pending Cronboard: A terminal-based dashboard for managing cron jobs github.com
15 pending 'Incoherent': Hegseth's Anthropic ultimatum confounds AI policymakers www.politico.com
16 pending Specs Should Be Equations, Not Essays fromanengineersight.substack.com
17 pending Stripe closed my account – no notice – my LLC was registered using Stripe Atlas self
18 pending Stripe closed my account – no notice – my LLC was registered using Stripe Atlas self
19 pending Musk touts California robotaxis but Tesla does nothing to get permits finance.yahoo.com
20 pending What does " 2>&1 " mean? stackoverflow.com
21 pending Increased urination urgency facilitates impulse control in unrelated domains pubmed.ncbi.nlm.nih.gov
22 pending Attorney General Finds Amazon Price Fixing, Urges Halt of Illegal Conduct oag.ca.gov
23 pending Elon Musk threatens to halt Tesla Giga Berlin expansion over union vote electrek.co
24 pending Kansas invalidates driver's licenses, birth certificates for ~1k transgender www.reuters.com
25 pending Show HN: Transcribe-Critic – Merge transcript sources for stronger transcript github.com
26 pending Show HN: I stopped building apps for people. Now I make CLI tools for agents github.com
27 pending Show HN: Decoy – A native Mac app for mocking HTTP endpoints locally decoy-app.com
28 pending Show HN: Smplogs – Local-first AWS Cloudwatch log analyzer via WASM www.smplogs.com
29 pending Draining wetlands produces substantial emissions in the Canadian Prairies theconversation.com
30 pending Show HN: I built a local AI-powered Ouija board with a fine-tuned 3B model github.com
31 pending Show HN: Protection Against Zero-Day Cyber Attacks self
32 pending Show HN: Librarian – Cut token costs by up to 85% for LangGraph and OpenClaw uselibrarian.dev
33 pending Show HN: Browser-based .NET IDE with visual designer, NuGet packages, code share xaml.io
34 pending Show HN: Batchling – save 50% off any GenAI requests in two lines of code github.com
35 pending Show HN: The best agent orchestrator is a 500-line Markdown file github.com
36 pending Show HN: Conjure – 3D printed objects from text description only conjure.tech
37 pending Show HN: I built a managed Claude AI and hosting service codedoc.us
38 pending Show HN: I made a directory for Claude skills skillsplayground.com
39 pending Show HN: Duck Talk – Real-time voice interface to talk to your Claude Code github.com
40 pending America Chose Not to Hold the Powerful to Account www.theatlantic.com
41 pending I built a 151k-node GraphRAG swarm that autonomously invents SDG solutions self
42 pending Ralph Wiggum Explained: Stop Telling AI What You Want – Tell It What Blocks You platform.uno
43 pending Show HN: Relay – SMS API for developers (send your first text in 2 min) self
44 pending You Just Need Postgres youjustneedpostgres.com
45 pending Show HN: A minimal Claude Code clone written in Rust github.com
46 pending Show HN: SAIA – SCUMM for AI Agents github.com
47 pending Palm OS User Interface Guidelines [pdf] cs.uml.edu
48 pending Child-free 'Disney adults' are transforming the company's theme parks www.businessinsider.com
49 pending How AI skills are quietly automating my workday medium.com
50 pending The Pentagon Feuding with an AI Company Is a Bad Sign foreignpolicy.com
51 pending Emacs Is a Lisp Runtime in C, Not an Editor thecloudlet.github.io
52 pending Snakes.run: rendering 100M pixels a second over SSH eieio.games
53 pending Secure Snake Home (SSH) snake.eieio.games
54 pending Making WebAssembly a first-class language on the Web hacks.mozilla.org
55 pending Will vibe coding end like the maker movement? read.technically.dev
56 pending Show HN: NotBuiltYet– Open-source library of civilisation problems worth solving shivankar-madaan.github.io
57 pending Rule of Three (Computer Programming) en.wikipedia.org
58 pending Show HN: Gonzales – Self-hosted internet speed monitor with Home Assistant github.com
59 pending Show HN: I'm building TaskWeave, a task orchestrator github.com
60 pending Apple Launch on Monday twitter.com
61 pending Larry Summers to resign from Harvard over Epstein ties www.reuters.com
62 pending In 2025, Meta paid an effective federal tax rate of 3.5% bsky.app
63 pending Model Collapse Ends AI Hype www.youtube.com
64 pending US role as global talent hub in doubt amid Donald Trump's visa crackdown www.ft.com
65 pending Show HN: I built a 50ms SPF record and Shadow IT scanner spf1.com
66 pending Show HN: Coding agents find the right GPU bottleneck 70% of the time, fix it 30% ayushnangia.github.io
67 pending Story of XZ Backdoor [video] www.youtube.com
68 pending Anthropic gives Opus 3 exit interview, "retirement" blog www.anthropic.com
69 pending Hubble could re-enter atmosphere as early as 2028 www.theregister.com
70 pending Long Range E-Bike jacquesmattheij.com
71 pending Burger King will use AI to check if employees say 'please' and 'thank you' www.theverge.com
72 pending Show HN: Riverse – persistent AI memory that grows with you, no RAG github.com
73 pending Anthropic ditches its core safety promise www.cnn.com
74 pending CIA offering Iranians instructions for using Tor xcancel.com
75 pending Fentanyl makeover: Core structural redesign could lead to safer pain medications www.scripps.edu
76 pending Say goodbye to budget PCs and smartphones – memory is too expensive now www.theregister.com
77 pending Linux 7.0 is coming: What to expect from the next major kernel release www.linuxjournal.com
78 pending Number of UK workers on zero-hours contracts hits record high ahead of crackdown www.bbc.co.uk
79 pending Show HN: Agent Swarm – Multi-agent self-learning teams (OSS) github.com
80 pending Show HN: One grammar, 18 YAML parsers – a Futamura projector in Common Lisp github.com
81 pending Nihilistic Violent Extremism en.wikipedia.org
82 pending Claude Code Bug triggers Rate limits without usage self
83 pending Men in their 50s may be aging faster due to toxic 'forever chemicals' www.cnn.com
84 pending Show HN: I built this toolbox with AI – never wrote a line myself tool.hikun.me
85 pending You Want to Visit the UK? You Better Have a Google Play or App Store Account www.heltweg.org
86 pending Comparing manual vs. AI requirements gathering: 2 sentences vs. 127-point spec self
87 pending Show HN: Parallel rsync launcher with fancy progress bars github.com
88 pending I Hate Trump's Awful Policies, but I Love That He's an Asshole www.mcsweeneys.net
89 pending Show HN: PyMOL-RS – Rust reimplementation of PyMOL with modern rendering github.com
90 pending Show HN: I built an AI that turns emailed PDFs into ledger entries in 60s baguno.app
91 pending Show HN: Better Hub – A better GitHub experience www.better-hub.com
92 pending Technical Excellence Is Not Enough raccoon.land
93 pending Show HN: Codex builds a working NES Emulator in one hour github.com
94 pending There is no reason Canadian Tire company should have any of my data infosec.exchange
95 pending Show HN: Skillscape – Engineering skills matrix without the spreadsheet www.skillscape.dev
96 pending Mumsnet campaign demands ban on social media for under-16s www.theguardian.com
97 pending Earth's heat to power 10k homes in renewable energy first for UK www.bbc.co.uk
98 pending Show HN: Nullroom.io – Experimental, stateless P2P messaging and file sharing www.nullroom.io
99 pending I don't know how you get here from "predict the next word." www.grumpy-economist.com
100 pending Giatse www.theguardian.com
Operations
Cron interval Every 5 minutes
Score refresh Every 5 min (updates) · 48h sweep every 10 min
Eval today 1319 evals · 9.4M in · 5.2M out
Total usage 1528 evals · 10.9M in · 6.2M out
Queue 0 evaluating · 100 queued/pending
Crawl surface top · new · best · ask · show (5 lists)
Architecture Cron v5 → Queue → Consumer (fan-out) · KV content cache · KV DCP cache
Models
Multi-model comparison, agreement analysis, and score distributions.
About HRCB | By Right | HN Guidelines | HN FAQ | Source | UDHR | RSS
build d633cd0+ahgg · deployed 2026-02-26 22:27 UTC · evaluated 2026-02-26 22:10:52 UTC