Qwen needs Python, Antigravity's demo was in ChEMBL, DeltaNet beats Transformers
Every URL the pipeline pulled into ranking for this issue — primary sources plus the supporting and contradicting findings each Researcher returned. Inline citations in the issue point back here.
Sources
Qwen 3.8 27B on my laptop can interpret model weights, too simonwillison.net
Yesterday’s hot paper was Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights by Rikiya Takehi, Jayden Teoh, Alex Zhang, and Omar Khattab. The paper (OK, the authors classify is as a “blog post”, but it’s paper-shaped) found that frontier models, in particular GPT-6 Astra, can both interpret and modify the raw weights of a machine learning model - literally a text description of the model containing a bunch of floating point numbers. I’ve been playing with…
Agentic cheminformatics with Google Antigravity deepmind.google
Learn how Google Antigravity and our Cheminformatics Skills accelerate drug discovery workflows, search compounds, and propose novel rigidified macrocycles.
StateBench: Testing how language models track state changes over long documents huggingface.co
CWE-bench v2: More coverage, quality, and transparency huggingface.co
CWE-bench v2 updates the vulnerability-detection benchmark with broader CWE coverage, cleaner ground-truth labels, and fuller documentation of how samples were sourced. The revision targets reproducibility gaps that let earlier code-security evaluations overstate model performance on real-world bug classes.
Palette: teaching a small model to build decks in your design language huggingface.co
Palette, from IBM Research, fine-tunes a compact model to generate slide decks in a user’s own design language rather than generic templates. The approach learns layout, typography, and color conventions from example decks so outputs stay on-brand without manual restyling.
References
Bio-IT World — Derek Lowe interview bio-itworld.com
AI excels at well-structured problems like protein folding but still struggles with the messy reality of small-molecule discovery; the primary causes of drug failure — target selection and human toxicity — remain outside the current reach of AI models.
Isomorphic Labs — IsoDDE announcement isomorphiclabs.com
The Isomorphic Labs Drug Design Engine more than doubles the accuracy of AlphaFold 3 on novel protein–ligand systems in the internal Runs N’ Poses benchmark, outperforming gold-standard physics-based methods for binding-affinity prediction.
IntuitionLabs — Pharma LLM security report intuitionlabs.ai
Indirect prompt injection via a malicious database entry can hijack an agent’s reasoning process; 83% of pharmaceutical organizations lack basic automated safeguards against AI-related data breaches.
Vibe Leaderboard — DrugDiscoveryBench 2026 vibeleaderboard.ai
On 82 expert-curated tasks, GPT-5.5 led at 51.6%, Gemini 3.5 Flash reached 50.0%, and Claude Opus 4.8 scored 47.2% — no single model dominates, and a superior agent harness can swing results by over 10 percentage points.
GlobAI — independent coverage of the Antigravity launch globai.org
Experts emphasize the distinction between a worked demonstration and a validated drug result; changing a single atom can redefine a compound’s properties, and a guessing model can quietly ruin a multi-million-dollar analysis.
Chenab Times — DeepMind cheminformatics coverage thechenabtimes.com
Antigravity unifies IUPAC names, Markush structures and SDF files, and allows export in industry-standard formats so outputs feed directly into existing lab pipelines — but DeepMind reiterates that proposed macrocycles remain testable hypotheses requiring wet-lab validation.
Merrill, Petty & Sabharwal — ‘The Illusion of State in State-Space Models’ (arXiv 2404.08819) arxiv.org
Despite their recurrent formulation, common SSMs like S4 and Mamba are expressively confined to TC⁰ and provably cannot solve NC¹-complete word problems such as S₅ permutation composition.
NVIDIA Research — ‘Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention’ research.nvidia.com
The original Gated DeltaNet tied erase and write to a single scalar gate; decoupling them improves multi-key retrieval and needle-in-a-haystack performance.
DeltaProduct paper (arXiv 2502.10297) arxiv.org
DeltaProduct uses Householder products to achieve higher-rank state transitions, overcoming the rank-1 limitation of the original delta rule and stabilizing effective rank as sequences grow.
SuperpowerDaily — independent coverage superpowerdaily.com
StateBench fills a ‘measurement gap,’ since traditional needle-in-a-haystack tests ignore the procedural complexity of enterprise state tracking; frontier models fall to 13% or less on the longest, hardest tracking tasks.
llm-stats.com — RULER benchmark summary llm-stats.com
RULER’s 13 tasks across retrieval, multi-hop tracing, aggregation, and QA consistently show ‘effective context’ is far smaller than claimed context, with 128K-window models cliffing at 32K–64K on non-lookup tasks.
Sebastian Raschka — ‘LLMs from Scratch’ DeltaNet chapter sebastianraschka.com
Achieving perfect generalization on Parity-like state tracking requires β_t ∈ (0, 2) to permit negative eigenvalues; missing this parameterization yields near-random accuracy.
R&D World — summary of Honnibal/Goedecke critiques rdworldonline.com
Critics argue the headline Qwen repair required a Python sandbox and a budget of 50 forward passes on practice problems — closer to an automated ML engineer than a ‘native’ weights reader; the ‘no-tools’ setting only held for toy 17K-weight networks.
Sean Goedecke, ‘LLMs are not superhuman…’ (via daily.dev) daily.dev
A three-line numpy snippet computing L2 norms of first-layer weights produces heatmaps that make the game obvious to any human — a paddle row for Breakout, a maze for Pac-Man — so identifying the game from labeled-channel toy weights is ‘relatively trivial’.
HuggingFace repo RikiyaT/Qwen3.5-0.8B-weight-space-programming huggingface.co
Independent reproduction from released bit-exact diffs lifts MATH-500 from 7.4% → ~50% and GSM8K from 13.8% → 67.1%; the edit modifies 131,072 entries across down_proj columns in layers 15 and 19, and cuts the token-limit termination rate from 93% to ~33%.
CryptoBriefing — Gemma 3 negative result cryptobriefing.com
Applied to Gemma 3 270M, which does not exhibit the Qwen-style repetition loop, the same weight-editing approach yielded negligible math gains — the authors concede the method ‘removes a barrier’ rather than ‘adding an ability.’
Coral Flavor — abliteration / steering prior art coralflavor.com
Abliteration uses weight orthogonalization to bake a contrastive activation direction permanently into weights — a lineage running through Arditi et al. (2024), Turner’s ActAdd, and Rimsky’s CAA; critics say Astra’s Qwen ‘fix’ is a specific application of this well-known steering-vector technique.
Hacker News discussion (item 50355251) news.ycombinator.com
Commenters flagged the paper-shaped blog post as bypassing peer review while adopting formal paper aesthetics, and noted that Astra’s success hinges on a 50-pass interactive search budget — low-budget variants show much smaller gains.