Microsoft tunes Qwen to Opus, Ai2 tops Google satellite AI, OpenAI runs Palantir
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
Turning Database Grading into RL Rewards: Post-Training Agents with ThinkingBox huggingface.co
OlmoEarth v1.3: State-of-the-art embeddings for the whole planet huggingface.co
What FDE means at OpenAI openai.com
What forward deployed engineering means at OpenAI
Ensemble forecasting with IBM Granite time-series models huggingface.co
IBM Research published a Hugging Face walkthrough for combining its Granite time-series models into ensembles, aimed at sharpening forecast accuracy on noisy real-world signals. The post details how to blend predictions from multiple Granite variants rather than relying on a single checkpoint.
Designing an AI serving harness built to last technologyreview.com
Enterprise differentiation is moving away from picking the biggest model and toward the serving stack wrapped around it, the MIT Tech Review piece argues. Years of chasing bigger context windows and benchmark scores have commoditized raw capability, leaving infrastructure as the durable moat.
AI’s hidden storage bill: What inference leaves behind technologyreview.com
GPU racks dominate AI cost conversations, but the back-end data generated by billions of daily prompts, images, and videos quietly piles up storage charges that executives overlook. The piece reframes inference economics around what happens to data after the model responds.
Claude Mods: Flipping the Harness Inside Out latent.space
Anthropic’s Claude Mods is pitched as an early ‘Attention Interface’ that inverts the usual agent harness: instead of wrapping the model to talk to tools, the harness wraps the user, surfacing what the agent needs humans to see and decide.
DeepSeek released DeepInfer, an open-source inference stack it says runs 155T tokens per day per cluster and generates $22M daily at roughly 80% margin, implying an 81-day hardware payback. The drop reframes the economics of serving frontier-class models.
References
ai.miraheze.org wiki entry on ThinkingBox ai.miraheze.org
Critics argue this ‘harsh’ design can overstate reliability deficits by treating an agent that succeeds 19 out of 20 times the same as one that fails every trial… tying evaluation to a specific JSON-RPC communication standard, potentially advantaging models whose training data included heavy exposure to MCP-style schemas.
CryptoBriefing — ThinkingBox vs tau-bench / SWE-bench comparison cryptobriefing.com
On ThinkingBox-Bench, the top-tier GPT-5.4 model achieved a 65.36% success rate on the first try (pass@1) but only a 25.25% rate when required to succeed 20 times in a row (pass^20)… 77.5% of failed traces in these environments are due to unrecovered tool errors rather than poor final answers.
ExplainX — Microsoft ThinkingBox stateful agent benchmark analysis explainx.ai
A 90% success rate sounds sufficient for a pilot, but for a recurring business process like insurance claims or automated banking, a 10% failure rate becomes a compounding liability… when agents move through 10+ steps, a small 5% error rate per step can cause the entire workflow success rate to plummet to 20%.
OpenTrain AI — GRPO post-training reproduction audit opentrain.ai
A primary bottleneck is that the private training pool used to achieve reported gains—such as boosting Qwen3.8-27B success rates by 9%—is not included in the public release. Furthermore, the repository lacks a standardized Docker setup or a robust CI pipeline.
Lucek.ai — RLVR critique on specification gaming lucek.ai
An agent may learn to satisfy the verifier without solving the underlying task; for instance, a coding agent might hard-code a successful health endpoint response to ‘pass’ a test while leaving the service broken.
Stanford CS329T lecture notes on RL for agents web.stanford.edu
Relative scoring methods like Group Relative Policy Optimization (GRPO) can inadvertently hide reward hacking because they only reward a model for being better than its siblings, not for being ‘correct’ in an absolute sense… current RL methods may merely reweight reasoning paths already present in a base model’s distribution rather than discovering fundamentally new logic.
Microsoft Research — ‘No One Knows the State-of-the-Art in Geospatial Foundation Models’ microsoft.com
In a survey of 152 papers, investigators found 46 instances where the same model, benchmark, and protocol yielded performance scores that differed by at least 10 points… 39% of geospatial foundation model papers fail to release model weights under any license.
Milan Janosov, Substack — ‘The Public Geospatial Foundation’ milanjanosov.substack.com
OlmoEarth and Tessera are blueprints for the community rather than mere ‘press release’ models; AEF embeddings perform comparably when used as frozen feature extractors, but OlmoEarth outperforms AEF substantially once fine-tuning is applied.
Ai2 blog — PLANTEUR benchmark allenai.org
On the leave-one-island-out measure, OlmoEarth v1.3 achieved a Mean Intersection over Union (mIoU) of 24.9, notably higher than Tessera v2 (19.1) and AEF (15.8).
allenai/olmoearth_pretrain GitHub — Artifact License github.com
Use is prohibited for weapons development, intelligence gathering, human surveillance, or policing, and for activities involving the removal of raw materials from the earth, such as mining or large-scale resource extraction.
ECCV 2024 — ‘Towards Latent Masked Image Modeling’ (Wei & Gupta) eccv.ecva.net
Joint optimization of online and target encoders in Latent MIM frequently leads to representation collapse, in which the model produces trivial, uniform outputs regardless of the input, as the encoder and target ‘chase’ each other into a low-variance subspace.
Ai2 blog — OlmoEarth deployment partners allenai.org
IFPRI uses the platform to generate county-scale crop maps in Kenya; the Amazon Conservation Association tracks deforestation drivers in Peru’s Madre de Dios; Global Mangrove Watch fine-tunes models to identify mangrove loss with higher precision than traditional specialist models.
fdeinterviewprep.com fdeinterviewprep.com
FDEs risk becoming ‘glorified field service techs’ who spend more time on customer complaints and bespoke patches than on core product innovation; employees describe ‘brutal’ 80-hour weeks and the potential to be ‘boxed in’ to customer-facing work.
explainx.ai — Colin Jarvis deployment-gap profile explainx.ai
OpenAI is ‘cloning’ Palantir’s two-decade-old playbook… a move toward ‘sticky’ 40–60% margins compared to the lower margins of commodity API licensing, with Jarvis (a former Palantir executive) importing the Delta/FDE model wholesale.
Tom’s Hardware — OpenAI/Synopsys GPT-Synopsys tomshardware.com
GPT-Synopsys is engineered to not only suggest code but to execute tools, analyze timing slack, and iteratively adjust physical layouts to meet power, performance, and area (PPA) targets.
industryaidecision.com — chip verification sign-off industryaidecision.com
Traditional deterministic EDA tools remain the ‘ground truth’ for checking the underlying physics of a design; AI cannot replace the high-fidelity validation required before fabrication.
Sierra AI — τ-Voice benchmark sierra.ai
In the τ-Voice benchmark, OpenAI models scored as low as 28% in telecommunications tasks compared to 71% in retail; real-world telephony conditions (8kHz compression, frame drops) can reduce task success rates by up to 20% versus ‘clean’ lab environments.
synthszr.com — Big Bench Audio synthszr.com
Initial evaluations of models like GPT-4o revealed a ‘reasoning gap’—over 90% accuracy on text-based logic dropped to roughly 66% when identical tasks were presented as speech.