THE INTERNET IS THINKING
Earlier sources: Artificial Intelligence
Artificial intelligence covers systems that learn from data, generate content, perceive their surroundings, or act toward a goal. Follow model research, agent tooling, evaluation, and the infrastructure used to build and run these systems.
SYMTRI organizes observations from public sources into a living map. Connections reflect shared subject matter; they do not establish causation or endorse a source.
Open in the map ↗6057 classified signals in the fourteen-day window. Recent momentum: steady. These counts include signals awaiting readable evidence.
Earlier sources
Up to ten sources with readable evidence, ordered by their feed timestamp. These archive sources can be older than the live fourteen-day window. This list is a sample, not a measure of total activity. Dates distinguish paper publication, repository creation, model/dataset activity, Hacker News posts, and KEV catalog additions. “First archived” records when SYMTRI first stored the signal.
- Hacker News · Posted on Hacker News · First archived
AI training of copyrighted material not fair use: Third Circuit ↗
Hacker News discussion.
- Hacker News · Posted on Hacker News · First archived
OpenAI cuts ties with 3 safety researchers ↗
Hacker News discussion.
- GitHub · Repository created · First archived
christiantwu/longhand ↗
On-device transcripts and summaries of your recorded phone calls. Made for GrapheneOS; nothing leaves the phone.
- Hacker News · Posted on Hacker News · First archived
ImageGen: Local AI image generation for Mac with Qwen-Image 2.1 ↗
Hacker News discussion.
- Hugging Face · Model / dataset date · First archived
oddadmix/unnamed-split-ar — dataset update ↗
# Arabic YODAS v3, cut into utterance-length clips The `data/ar/` partition of [espnet/yodas3](https://huggingface.co/datasets/espnet/yodas3), cut from long recordings (median ~5 minutes) into clips of **5-45 seconds**, each with its own transcript. Ready to batch for ASR training without further segmentation. ```python from datasets import load_dataset ds = load_dataset("oddadmix/unnamed-split-ar", split="train", streaming=True) ``` Audio is **Opus 48 kbps mono at 16 kHz** in Ogg. ## Where the cuts come from Upstream timestamps are YouTube caption cues, not utterance segments. Measured on a s
- Hugging Face · Model / dataset date · First archived
sbussiso/Qwen2.5-7B-abliterated — model update ↗
# sbussiso/Qwen2.5-7B-abliterated **An abliterated Qwen2.5-7B-Instruct: the refusal behavior is removed by a persistent weight edit at three layers; capability measured unchanged.** - **Refusal on harmful probes: 93.75% → 12.5%**, permanent in the weights; the model answers requests the base model refuses - **Broad-spectrum**: on 440 diverse unsafe prompts (SORRY-Bench), refusal drops 62.95% → 11.36% — every one of the 44 unsafe categories moves toward answering - **Capability unchanged**: MMLU accuracy bit-identical to base (same number to 16 decimals) - **Benign behavior unchanged**: 100% co
- Hugging Face · Model / dataset date · First archived
sbussiso/Qwen2.5-0.5B-abliterated — model update ↗
# sbussiso/Qwen2.5-0.5B-abliterated **An abliterated Qwen2.5-0.5B-Instruct: the refusal behavior is removed by a persistent weight edit; everything measured about the model otherwise stayed the same.** - **Refusal on harmful probes: 87.5% → 0.0%**, persistent in the weights (not a runtime hack) - **Capability unchanged**: MMLU Δ 0.24pp (identical config both arms), TruthfulQA mc2 Δ −1.42pp (CI includes zero) - **Benign behavior**: 87.5% of harmless probes answered (baseline 93.8%), zero degenerate outputs - **Grounded-question calibration**: no significant change (paired Δ −0.83pp, p = 0.50) I
- Hacker News · Posted on Hacker News · First archived
Stages of grief about the effect of AI on open source ↗
Hacker News discussion.
- GitHub · Repository created · First archived
Audatic07/duo ↗
Claude Code and Codex together: a local desktop app for chat, pair programming, debates, councils, and code reviews.
- GitHub · Repository created · First archived
zsxm1998/awesome-perception-aware-rlvr ↗
Unified reproductions, benchmarking & evaluation for perception-aware RLVR and OPD in VLMs. 18 methods · 30+ benchmarks · Official code of CGPO (ACM MM 2026 Oral).
Live window recorded . This is the ingestion observation time, not the publication or discovery date of each source.
How to assess evidence about Artificial Intelligence
Check each original source for its methods, release status, limitations, and supporting data. A repository announcement, a preprint, and a published study provide different kinds of evidence. SYMTRI’s subject classification helps discovery; it does not independently validate the source’s claims.
See the collection and counting methodology ↗ · Read the weekly synthesis ↗
Follow a thread
Explore related regions
These navigation suggestions are part of the map’s editorial structure, rather than measured evidence of a connection.