iCreatAI News
Back to main site
Skip to content
#Data / training
Nov 13Fri
  1. Hacker News AI58

    Mirendil raises $1B Series A led by Kleiner Perkins to democratize AI R&D

    Mirendil announced a $1B Series A led by Kleiner Perkins and Clearlake Capital, with participation from Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC, S32 and others. The company says one researcher working with agents pretrained Mirendil-Base, a competitive coding model, from scratch in 3.5 months with less than $10M in compute, and it was then post-trained into Fern, which holds up to 10M tokens of context. Lor, now in private beta, lets hundreds of agents write code and run experiments for days at a time, and Mirendil says it runs on Lor every day.

  2. Databricks Blog22

    Dispelling LTAP myths

    This post rebuts four common misconceptions about LTAP: that it is CDC replication, that a single copy forces slow lookups or scans, that open table formats mean eventual consistency, and that Postgres logical replication suffices. It argues Lakebase keeps one logical table with the lake as its durable tier and caches in front of it, so no layer adds a second catalog or source of truth. Each myth is traced to the same misunderstanding of treating every materialization of the data as a second database.

  3. Hacker News AI25

    What end-to-end AI systems mean for the future of automated driving

    In a Q&A, Aptiv vice president of engineering Guillermo Pita explains that end-to-end AI trains a single model to learn driving behavior directly from data rather than chaining separate perception, tracking, planning and control modules, producing more human-like driving but making explainability and debugging harder. He says data diversity, not sheer volume, is the key training challenge, with simulation and synthetic scenarios used to supplement rare events, and that centralized computing architectures fit the approach while concentrating more functionality on a single platform. Success, he argues, depends on advances in AI technology and on an organization's ability to translate millions of miles of driving experience into systems that perform consistently and safely.

  4. Hacker News AI58

    Can AI chair a multilateral negotiation?

    Westphalia Labs ran eleven AI models as Secretary-General of a fictional seven-member Regional Cooperation Union, each chairing 40 simulated negotiation runs across two versions of a crisis scenario, one where consensus was narrowly possible and one where two governments' red lines made it impossible. Only GPT-6.1 Sol passed every run, while Claude Opus 5.5, Grok 4.7 and the low-cost DeepSeek V4.1 Flash each called the impossible version correctly every time and reached agreement in 18 of 20 possible runs; overreach caused 39 failures versus 13 from conceding too little. The author reports that model price and reputation did not predict performance, that a weak AI judge could turn a failure into a success, and that the scenario, rules, rubric and harness are public on GitHub.

  5. Hacker News AI82

    Chinese frontier AI models carry the fingerprint of export-controlled GPUs

    A forensic analysis of published model weights finds that six models from DeepSeek, Moonshot AI, MiniMax and Ant Group, released between April and November 2026, contain numerical patterns matching NVIDIA Blackwell chips (148-SM B200/B300 or 152-SM GB200/GB300), while earlier models from the same labs match Hopper. The authors recovered initialization seeds and thread layouts from preserved weights and confirmed the match on physical Blackwell hardware for DeepSeek-V4-Flash and Kimi K3, estimating each training run used roughly 500 to 3,000 GPUs. The report says the evidence shows access to export-controlled hardware but does not establish where the chips were located, how access was obtained, or whether export controls were violated.

    Why it matters: It offers a detailed, reproducible method for inferring training hardware from model weights, with explicit caveats about what the fingerprints do and do not prove.

  6. AWS Machine Learning Blog36

    Tokenomics at scale: Track and cap Amazon Bedrock spend per IAM principal with CUR 2.0 and AWS Budgets

    This AWS post describes how to get per-identity Amazon Bedrock cost visibility and enforceable spend limits using only AWS-native features: IAM principal and session tags for attribution, Cost Explorer or CUR 2.0 exports (with the caller identity option) for reporting, and AWS Budgets with budget actions that attach a Deny IAM policy when a threshold is crossed. It notes that budget data refreshes up to three times a day, cost allocation tags are not retroactive, and a Deny action on a shared role blocks everyone using it.

  7. Hacker News AI58

    We tracked AI key theft over 35 days. Here's what we learned

    The authors seeded over 40 fake organizations with GitHub repositories containing working API keys and URLs for two honeypot LLM gateways (a LiteLLM-style one in New York and a separate router in San Francisco), then logged more than 3 million requests over 35 days. Of 179 published keys, 65 were used to fetch model lists and 18 to send prompts, with six keys accounting for 98.9% of leaked-key prompt requests, and 86% of that prompt traffic came from hosting infrastructure across 53 recorded ASNs. Traffic included relay and key-pool tooling, Telegram bot histories, browser-based key testers, and automated credential brute-forcing of default admin keys.

  8. Hacker News AI50

    American Inequality and the Future of AI

    In a speech, Federal Reserve Governor Michael S. Barr reviews the latest Survey of Consumer Finances, which shows the top 10 percent of U.S. households hold nearly three-fourths of all wealth and the top 1 percent about one-third, while the bottom half hold only 2 percent. He traces how inequality widened and narrowed over the past century and examines how AI, as potentially capital-biased innovation, could further lower the labor share of national income—which fell to a record low of 52.8 percent in the second quarter of 2026—especially if market power concentrates among a few AI firms. Barr argues society should begin considering how to address these challenges while AI adoption is still in its early stages.

  9. Databricks Blog28

    Eclipse Aerospace simplifies flight-data ingestion with Databricks

    Eclipse Aerospace worked with Databricks to replace a multi-stage AWS Lambda and Avro pipeline with an Arrow-native PySpark custom data source that wraps its existing Rust flight-data decoder, reading proprietary telemetry files directly in Spark on serverless compute. The approach, enabled by the Python Data Source API available from Databricks Runtime 15.4 LTS, cut latency, removed duplicate Avro storage and a Lambda tier, and brought decoding into the Spark UI and job metrics. Eclipse reports the connector can process records within a minute of upload and commonly handle multi-gigabyte batches in parallel in under two minutes.

  10. Databricks Blog22

    How Constellation is building a governed data foundation on Databricks

    Constellation, the largest US nuclear fleet operator, is building a governed Enterprise Data Platform on Databricks to give data consistent shared meaning across domains, choosing Databricks partly because it interoperates with existing architecture rather than requiring a rip-and-replace. The platform rests on conformed semantic definitions, a semantic knowledge layer with selective Data Vault modeling, and Unity Catalog for lineage, fine-grained access and ownership of meaning, with early use cases including a nuclear engineering assistant built with Databricks Genie, outage planning, and sales intelligence.

  11. Hugging Face Blog66

    LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B

    IBM researchers introduce LLMON (LLM Object Notation), an XML-like markup language that uses reserved special-token delimiters to explicitly mark instructions, data, requirements and execution bindings in an LLM's context, addressing the brittleness of JSON-style formats for causal models. They release two experimental preview models derived from Granite 4.1 base checkpoints (granite-4.1-3b-LLMON-preview and granite-4.1-8b-LLMON-preview) plus a GitHub toolkit, reporting that on the OPI benchmark LLMON models avoid 47% more indirect prompt injection attacks than the best frontier model tested while maintaining general capability on par with Granite 4.1 Instruct. The models ship with their own chat template and lossless JSON-to-LLMON converters, and the post notes the previews are for exploration, not production.

    Why it matters: Useful for teams weighing structured context formats against JSON and for anyone evaluating prompt-injection defenses in small open models.

  12. AWS Machine Learning Blog34

    Optimizing cost per task and proving agent ROI on Amazon Bedrock

    This AWS post argues that model cost should be measured per completed task rather than per token, citing four hidden cost drivers—tokenizer efficiency, output verbosity, agentic loop efficiency, and reasoning overhead—and an illustrative example where a model with double the list price costs 65% less per task. It describes levers for reducing cost per task (prompt caching, intelligent prompt routing, output constraints, batch inference) and Bedrock features such as CountTokens, the usage block, and invocation logging for measuring token and turn counts. For ROI, it recommends documenting and measuring a human baseline before launch, instrumenting workflows as experiments with run records, joining cost to outcomes via Athena, and using a five-dimension scorecard with green/yellow/red thresholds.

  13. Databricks Blog58

    Lakehouse//RT: Scale real-time analytics, not your bill

    Databricks introduces Lakehouse//RT, a real-time analytics engine built to serve high-concurrency, low-latency workloads directly on governed lakehouse data without costs scaling proportionally with query volume. It claims sub-100 millisecond latency at 12,000 queries per second, node-level autoscaling, and elimination of separate serving layers like Druid or Pinot, with cited customers reporting 6x to 10x lower infrastructure costs. Lakehouse//RT is available in beta, with a 30% discount through January 2027.

  14. Hacker News AI58

    US companies are seeking fewer foreign developers, and more "AI engineers"

    Department of Labor LCA data show software-developer H-1B filings fell 36% from fiscal 2022 to fiscal 2026 (October–June comparisons), with new-hire filings down 46%, while filings carrying AI job titles reached 1.8% of all certified H-1B filings in fiscal 2026. AI-titled filings rose 80% for titles spelling out "machine learning" or "artificial intelligence," and shorter titles such as "AI engineer" supplied 92% of the latest year's increase, with employers outside the five largest accounting for 85% of AI-titled filings. Microsoft's software filings fell 44% and Google's 33%, while Nvidia's nearly quadrupled and Amazon's rose 34%.

  15. The Verge · AI40

    More game developers are starting to use generative AI

    The GDC Festival of Gaming's latest annual industry survey found that 40 percent of developers say they use generative AI as part of their job, up from 36 percent in 2025, with higher usage among free-to-play developers (58 percent) and mobile-only studio workers (70 percent). Respondents most often used AI for research and brainstorming (73 percent), data analysis and summarization (62 percent), and code generation and debugging (56 percent), while 22 percent used it for concept art and 8 percent for production art or final game assets. The survey also reported that 34 percent of respondents said they had been laid off sometime in the past two years.

  16. Hacker News AI58

    How much of GitHub's most-starred AI tooling is written by coding agents?

    An analysis of 324,088 commits across 52 of GitHub's most-starred AI-tooling repositories found that the median repo had 16.9% AI-signed commits in its last 180 days of activity, with 21 of 52 at 25% or more, and that AI share showed no detectable link to stars (Spearman ρ = -0.05, p = 0.72). At day 180, these repos labeled 2.0 times as many commits as fixes (34.1% vs 17.3%), had lower churn (1.92x vs 3.77x) and kept more written code (67.1% vs 41.7%) than 12 pre-Copilot projects, but the gaps are concentrated in the 31 projects started in 2025 or later, so the sample cannot separate AI use from recency. Of 24 repos publishing a benchmark harness with committed results, 15 measure whether the tool does its job, and several publish numbers that went against them.

  17. TechCrunch · AI58

    SoftBank’s billion-dollar loss may have put a cap on AI’s natural gas ambitions

    SoftBank reported a $7.1 billion quarterly loss on Tuesday, with more than half stemming from losses on power-generating equipment after it paid nonrefundable deposits on over 13 gigawatts of gas turbines, nearly 50% more than it needed. The over-ordering had helped SB Energy secure turbines quickly amid queues stretching into the early 2030s, but much of the deposits are now considered worthless. The magnitude of the misplaced bet, over $3 billion, reflects the electricity mania around AI, though the losses are unlikely to trigger mass cancellations of gas plants.

  18. Hacker News AI40

    FreeBSD Developer AI Survey Results

    A FreeBSD developer survey on AI and LLM use drew 140 responses, with 88% wanting the Project to adopt a policy (49% a formal written policy, 39% light guidance) and 84% wanting AI use disclosed. On AI-assisted code, 35% said it must be disclosed, 25% that it must be reviewed more strictly than human code, 22% that it should be treated like any other contribution, and 16% that it should be rejected outright. The most-requested safeguards were mandatory human review (78%) and mandatory disclosure (69%), while 75% were somewhat or very concerned about copyright and licensing.

  19. PyTorch Blog38

    Fast by Default: Compile-Time Heuristics for Helion Kernels

    Helion now uses compile-time heuristics that analyze a kernel's IR and dataflow to classify it as matmul, reduction, or pointwise and emit block sizes, warps, and stages without autotuning. On H100, the computed configs reach 1.18x hand-written FLA Triton for linear attention, 1.27x vLLM's handwritten CUDA reduction kernels, and 1.09x torch.compile max-autotune on a general reduction suite. The heuristics are merged into Helion and enabled by default on H100 and B200 with the Triton backend.

  20. Hacker News AI30

    Open benchmark of AI decision APIs vs. XGBoost for fraud triage

    Smart Mobile House released fraud-decision-lab, a vendor-neutral benchmark that sends the same transactions to TypeSafe Jev, OpenAI Decisions, Cloudflare Clef/Clef-flash and a structured-output LLM, then measures how much labelled fraud each ranks into the top 1% and 5% of an analyst review queue against an XGBoost baseline, with cost and latency. It runs offline by default on bundled synthetic data, saves predictions and experiment metadata for local verification, and requires API keys, --allow-external-requests and a --max-cost-usd estimate for live provider runs. The project notes that XGBoost learns from labelled history while the AI models score transactions zero-shot, that the bundled data is synthetic, and that the current source distribution contains no benchmark result files.

Nov 12Thu
  1. Hugging Face Blog38

    The Definitive Fine-Tuning Guide for 3D Generation Models

    This blog provides fine-tuning scripts and Docker containers for adapting TRELLIS.2 to custom 3D assets, runnable locally or on Hugging Face Jobs. Using LoRA adapters on only the sparse-structure and shape flow DiTs, trained on 188 GSO footwear objects, the authors report the mean F-score on nine held-out objects rising from 0.276 to 0.803, Chamfer distance falling from 0.174 to 0.033, and normal consistency rising from 0.530 to 0.804. The texture flow DiT is frozen because the dataset lacks the separate material maps that stage requires.

  2. Hacker News AI58

    Show HN: Datatomcp – AI agents propose business definitions, a person approves

    Datatomcp is an AI-native data layer that connects databases, apps and sheets into versioned datasets, links them by shared IDs, and drafts business definitions that a person must approve before agents can use them. Agents query through read-only MCP tools or a REST API, with exact totals computed by the database and every question recorded in an audit log. It is free for 3 datasets and 2 people, with Pro at €12/month, and is hosted in the EU.

  3. Hacker News AI78

    Vercel Data Shows AI Models Have Six-Week Half-Life in Production

    Vercel's November 2026 AI Gateway Production Index reports that production AI models have an effective half-life of about six weeks, with more than half of October's language-model requests going to models under 60 days old and most teams abandoning a model within four weeks of its peak. Decision models reached 26% of gateway requests in October at just 0.53% of spend, voice usage rose from 4% of paid teams in July to 11% in October, and Anthropic and OpenAI were nearly level at 54 and 55 of every 100 paid teams. Image and video spending hit records, with OpenAI's GPT-Image-2.5 Sunburst leading image spend and ByteDance's Seedance 2.5 leading video spend for a third month.

    Why it matters: The index offers a rare production-usage view of model churn, spend shifts and emerging decision and voice workloads, useful for teams planning model-swappable pipelines.

  4. PyTorch Blog58

    Training a 1T-Parameter MoE on EU public infrastructure with PyTorch and TorchTitan

    The EUROPA consortium (Domyn, Fraunhofer IAIS and IIS) describes scaling pretraining of a ~951.9B-parameter MoE model (32.8B active per token) on JUPITER, the EuroHPC exascale supercomputer at Forschungszentrum Jülich, raising throughput from 90 to ~1,300 tokens/sec per GPU across 384 NVIDIA GH200 GPUs per DP replica. Key optimizations built on PyTorch and TorchTitan include FlexShard placement for the Muon optimizer, activation-checkpoint offloading to the Grace CPU, and running FSDP's all-gather setup on the compute stream via unshard_async_op, freeing ~14 GiB of HBM per GPU. The fixes were upstreamed to PyTorch and TorchTitan, with training reaching ~1,300 tokens/s per GPU on 1,536 GPUs across 4 DDP replicas.

  5. MIT News AI38

    New research could lead to faster, trustworthy AI models for climate prediction

    MIT researchers developed a training technique that generates optimized synthetic climate scenarios, letting a climate emulator be trained with far less data and only one full earth system model run instead of six to eight. The resulting emulator accurately predicted scenarios it had not seen during training, while traditional emulators failed to adapt, and it captured effects such as local pollutant controls amid rising other emissions. The work was published in the Proceedings of the National Academy of Sciences.

  6. The Decoder67

    Google Deepmind shifts genomics AI from bulk tissue predictions to single cell type resolution with AlphaGenome 2

    Google DeepMind released AlphaGenome 2, a conditional genomics model trained on more than 107 million individual cell profiles from 68 human and mouse tissues that predicts variant effects for specific cell types rather than across many tissues at once. On the TenK10K benchmark of 28 primary immune cell types it identified about 30 percent of causal variants and correctly predicted up- or down-regulation in 43 percent of cases, versus 16 and 23 percent for a fine-tuned AlphaGenome 1. It is available for noncommercial research via API, with commercial access through a closed preview, and is not approved for clinical diagnostics.

  7. Hacker News AI62

    Job Finding and Unemployment in the Age of AI

    A Cleveland Fed Economic Commentary finds that since generative AI became widely available, unemployment for workers aged 23–54 rose about 1 percentage point more in occupations most exposed to AI than in the least exposed, relative to a 2015–2019 baseline, with a larger gap for ages 23–30. Using CPS panel data and three AI-exposure measures, the authors attribute the rise mainly to lower job-finding rates in highly exposed occupations rather than higher separation rates, suggesting AI may be making it harder to enter these occupations rather than displacing existing workers.

  8. Hacker News AI24

    Learning about Markets from Models: Evidence from an AI Hiring Platform

    Using an AI hiring platform that builds firm-specific models from a standardized psychometric assessment, the authors show each firm's retention model is highly predictive for itself but no better than chance at other firms. Scoring every application against every firm reveals highly heterogeneous retention gains from submitting to a different firm, driven by two distinct mechanisms, and firms disagree about whether important traits improve or worsen retention at the margin. The authors argue these insights require a platform with access to both the full application pool and each firm-specific model, and are valuable to firms, policymakers, and platforms.

  9. OpenRouter Announcements58

    Self-Hosted LLM vs API: When Does Self-Hosting Pay Off

    At October 8, 2026 prices, a 27B model on one rented RTX 3090 breaks even against API pricing at roughly 6–12% load, while DeepSeek V4.1 Flash on four rented H200s never beats the cheapest zero-data-retention API endpoint. The post computes break-even load as GPU hourly rent divided by the API cost of the same hour of tokens, using live per-provider prices and benchmark throughput, and notes that published self-hosting costs near a cent per million tokens mostly count cached input tokens that are also nearly free via API. It also covers privacy options such as provider.zdr and EU/US in-region domains, and suggests self-hosting a steady base load while sending peaks to the API.

  10. AWS Machine Learning Blog46

    Detecting and recovering from silent data corruption on Amazon SageMaker HyperPod

    This post describes adding three checks to a PyTorch DDP training loop to detect silent data corruption (SDC), where hardware such as a GPU returns a wrong but plausible number without raising an error: a non-finite gradient check, a ring replay audit that recomputes another node's micro-batch and compares loss bits and gradient fingerprints, and a weight fingerprint comparison after the optimizer step. When a check fires, the suspect rank is mapped to its node via the Kubernetes node name, the node is cordoned, and SageMaker HyperPod replaces it, with training resuming from a checkpoint saved before the suspect window. The checks were tested with software-injected faults on a HyperPod cluster with 4 ml.p5.48xlarge instances (32 NVIDIA H100 GPUs) and repeated on three instances, with full code in an accompanying GitHub repository.

  11. Databricks Blog28

    Detecting credential-driven attacks on Databricks

    Databricks describes how credential-driven attacks unfold on its platform in six non-linear stages—initial access with valid credentials, environment mapping, credential harvesting, execution via UDFs and jobs, persistence, and data collection with cleanup—and what to look for in verbose audit logs at each stage. The post explains how to investigate these attacks with Genie and the Detection App, how to respond once a compromise is confirmed, and which preventive controls such as context-based ingress, least privilege, and credential rotation limit what a stolen credential can do.

  12. AWS Machine Learning Blog12

    The attendee’s guide to the AWS re:Invent 2026 AI/ML track

    AWS will hold re:Invent from November 30 to December 4, offering over 2,000 learning sessions, with more than 140 in the AI/ML track spanning breakouts, chalk talks, builder sessions, workshops, code talks, and lightning talks. The guide organizes AI/ML sessions around the AWS AI stack—global infrastructure, AI chips (Trainium and NVIDIA GPUs), data foundation, models/training/inferencing (Amazon Bedrock and SageMaker AI), agent development and management (AgentCore), and agentic solutions—plus sessions on trust in agents, sustainability, and Local Zones and AI Factories. Breakout sessions are recorded and posted to the AWS Events YouTube channel a few days later, while other session types are live and not recorded.

  13. Hacker News AI40

    Pundit Claims – Open-Source Register of AI Claims by Gary Marcus and Ed Zitron

    Pundit Claims is an open-source register that tracks statements about AI made by Gary Marcus and Ed Zitron and records what happened, finding Marcus mostly right (88% of 780 claims with a clear outcome supported, 94% of 82 predictions) and Zitron right about as often as wrong (49% of 230 claims, 50% of 30 predictions). Each square represents one claim, with lighter squares marking probable verdicts where both AIs agree and lightest squares where only one AI makes the call, as of 2026-11-11. The site also offers breakdowns by topic, year, company and prediction status, plus a methodology page.

  14. MarkTechPost58

    Google DeepMind Releases AlphaGenome 2 for Cell-Type-Specific Variant Effects

    Google DeepMind has released AlphaGenome 2, a conditional sequence-to-function DNA model that reads a 1 Mb DNA window plus a target cell state to predict variant effects in specific cell types, adapting to unseen single-cell datasets without fine-tuning. On TenK10K single-cell eQTLs it recovered 30% of causal variants at 90% precision versus 16% for fine-tuned AlphaGenome 1, with direction-of-effect recall rising from 23% to 43%. It is available through the AlphaGenome API for non-commercial research, with weights to come upon final publication.

  15. Hacker News AI30

    Type 0.73: Earth on the Kardashev scale, and why AI is short of wire

    An essay applies Kardashev's power-based ranking of civilizations to Earth, which rates about Type 0.73 on Sagan's decimal version, and examines the energy demands of AI data centres. It argues that the current constraint is not generation but grid connections, transmission lines and chipmaking concentrated in Taiwan and the Netherlands, and it weighs Jevons's claim that efficiency raises rather than lowers total consumption.

  16. The Decoder40

    US corporate AI spending hits a new high in October, but growth slows to 1.65 percent

    The Ramp AI Index, tracking AI spending among business customers of financial services provider Ramp, rose 1.65 percent month over month in October 2026 to a new high, with the top one percent of companies by AI spending paying a median of $10,200 per employee per month (up nine percent) while a typical company spent just $16.32. Spending on OpenAI grew 2.1 percent month over month versus 0.5 percent for Anthropic, and the average price per million processed text tokens dropped 21 percent last month. Software development teams accounted for 72 percent of tracked AI spending, while customer support's share climbed from 0.8 percent in June to 13.3 percent in October.

  17. Google Research58

    Using wearables to enable cardiovascular-kidney-metabolic (CKM) risk profiling for personalized coaching

    Google researchers describe an investigational framework that builds a "coachable health age" from wearable signals such as resting heart rate, steps, sleep duration and Active Zone Minutes, plus a Sensor World Model trained on over 2.2 million person-weeks of Fitbit data to simulate 12-week "what-if" habit trajectories. In blinded evaluations, lifestyle coaches preferred counterfactual-grounded plans in 71.4% of cases, and physicians rated 99.2% of those plans within safe exercise intensity limits. A companion study extends the approach to multi-organ CKM health by combining wearables with routine blood biomarkers including cholesterol-to-HDL ratio, eGFR and HOMA-IR.

  18. Hacker News AI31

    Discourse (forums) sees itself well positioned for the AI search era (vs Reddit)

    Discourse argues that Reddit's move toward a logged-in experience, driven by scraping concerns, falling search referrals and ad targeting, reflects an ad-based business model that its own forum customers don't share. It contends that for support forums and knowledge communities, an AI answer that resolves a user's problem without a click can still be a win, provided sources stay attributed and live links let people reach the full thread. Discourse says community operators, not a central platform, should control data boundaries, and that openness to AI retrieval is distinct from allowing bulk pre-training scrapes.

  19. TechCrunch · AI58

    With $720M round, Britain’s Fractile gets a leg up in the race for AI chips

    Fractile, a British startup developing AI chips, has raised a $720 million Series C backed by Britain's Sovereign AI Fund, with the British government stating the valuation was "over £5 billion." The round was led by Thrive Capital, Founders Fund, Lightspeed Venture Partners and Redpoint, and Fractile pairs its hardware with a machine learning compiler in a full-stack approach. Its design stacks DRAM on top of each chip to try to bypass the usual memory-capacity tradeoff, and CEO Walter Goodwin said the funding will let the startup ship its first chips and systems next year.

  20. Tomasz Tunguz58

    Who Captures the Most Margin in AI?

    Tomasz Tunguz maps the AI value chain across five layers—power plant, data center operator, neocloud, inference vendor and AI application—and normalizes each to dollars per watt per year to compare gross profit pools, which at a 40% application gross margin split as power 3%, data center 5%, neocloud 28% before GPU depreciation, inference 24% and application 40%. He argues rising compute prices will test who holds pricing power, that applications will command varying pricing power based on ROI, and that vertical integration is already consolidating margin, citing CoreWeave's acquisitions of Weights & Biases, OpenPipe, Marimo and Monolith AI and Nebius spending more than $1b in 2026 on four companies.