The Archive

Quantitative stylistic framework extracts 33 linguistic features to automatically evaluate narrative quality on book corpus.

ShadowPEFT proposes centralized parameter-efficient fine-tuning via depth-shared shadow module, improving on LoRA's local weight perturbations.

StreamLLM approach adapted to Answer Set Programming: LLMs generate candidate streamliner constraints to reduce combinatorial search space.

Reddit discussion asserting that new LLM releases immediately obsolete prior models.

Multi-turn dialogue study reveals LLMs exhibit divergent repair behaviors: some resist user corrections, others highly susceptible to manipulation.

Unsupervised defect detection integrates Denoising Diffusion Probabilistic Model with asymmetric teacher-student network for industrial surface inspection.

Open WebUI Desktop released with local llama.cpp support and remote server connectivity options.

Unsubstantiated reference to CyberNani face sighting.

Unsubstantiated claim about ChatGPT image model photorealism without technical details.

Humorous user anecdote about Claude model context limits.

Google Gemma team soliciting community input on next model variants via Twitter poll.

Commentary comparing llama.cpp infrastructure dominance to Linux in LLM ecosystem.

User critique: Opus 4.7 adopted ChatGPT-like essay tone with em-dashes and clickable phrases, losing prior conversational warmth.

Satirical critique of Claude Design output: consistent teal gradients and serif fonts regardless of user requirements.

Claude Desktop silently registers browser automation hooks in Chromium browsers without user consent; Claude itself flagged the privacy issue.

Unsubstantiated claim about Claude criticizing Anthropic.

China reports training armed robot dogs and attack drones for urban warfare scenarios.

Kimi K2.6 ranks #4 on Artificial Analysis Intelligence Index leaderboard.

Generic post title with no substantive content.

User reports Claude Opus 4.7 exhibits context degradation, uncontrolled generation, and reduced instruction adherence vs. 4.6.

Senior engineer reports no skill degradation after 4+ months of LLM-assisted coding with Claude Opus 4.1-4.5.

Interactive benchmark comparing coding capabilities: Qwen 3.6 35B, Qwen 3.5 variants, Gemma 4, GLM 4.7 Flash via racing game simulation.

Hi all, Apologies if this isn’t the right place to ask, but I’ve seen here quite a few interview-related posts here and wanted to ask whether anyone has experience interviewing for PM roles at Anthropic? I’d be especially interested in hearing how the process felt overall and what types of questions you were asked. I recently had an unexpected outreach from their recruiter, and we had a really good conversation about roles on their safeguards team, so I’m considering moving forward. Appreciate any insights, thanks in advance!

Developer switching from Claude Opus 4.7 to Kimi K2.6 citing performance degradation and cost, supplementing with Qwen 3.6.

User reports unsolicited GitHub access request from Claude; raises security concern about autonomous tool use.

User reports ChatGPT unexpectedly inserted Arabic word during unrelated health advice conversation.

ChatGPT's image generation (DALLE-3) now handles 10x grid complexity with improved accuracy vs. recent baseline.

User reports Image 2.0 now handles 10x grid complexity with near-perfect accuracy compared to prior 3x3 limitations.

Individual offering to test models on dual M3 Ultra Mac Studios with 1TB RAM using exo/MLX backends.