Models, products and AI startups, near-duplicates collapsed to the most credible source
Open-source freehire aggregates 3M+ tech jobsOpen-source project freehire searches 3M+ tech jobs from company career boards, combining a job tracker with AI-assisted resume tailoring; saved skills and preferences shape job matches and resume edits. It is available on GitHub for community stress testing.
JB Financial Group launches $22M AI investment fundJB Financial Group formed the "JB New Growth Investment Association No. 1," an AI strategic fund worth up to 30 billion won, backed by Jeonbuk Bank, Gwangju Bank and JB Woori Capital. Its first investment went to AI security firm Cypher Data, part of a strategy to tie investments to business cooperation across the group.
Lunit, Summit team up on AI biomarkers for cancer drugKorean medical AI firm Lunit will work with Summit Therapeutics, applying its Lunit Scope IO pathology platform to find biomarkers tied to treatment response for the dual-antibody drug ibonesimab, helping select likely responders. Summit is developing the PD-1/VEGF candidate globally.
llama.cpp discloses out-of-bounds write flawllama.cpp's Pocket-TTS component has a heap out-of-bounds write flaw triggered by a malicious mmproj file; it requires local access and a public exploit exists. Upgrading to build b11796 fixes it, but the root-cause patch is still open.
OpenAI works with academia for scientific understandingOpenAI shares model results on the Kakeya conjecture, Exponential Time Hypothesis, and quantum chromodynamics, and how researchers can verify and build on them.
SK Hynix plans about $12B special dividend on Kioxia stakeSK Hynix plans a special dividend of about 16 trillion won ($12 billion), distributing most of the 20 trillion won gain booked on its Kioxia stake in the first half, the DongA Ilbo reported. The prospect helped SK Hynix shares buck a broad AI stock selloff.
Nvidia Reportedly Eyes $2,000-per-GPU Yearly Software RevenueNvidia is reportedly looking to software revenue as a growth driver, with a figure of $2,000 per GPU per year cited from Huang. If realized, it would shift the company toward recurring revenue tied to its hardware.
Asia's busiest earnings week puts AI boom in focusAsia's busiest earnings week will test the AI trade as investors look for proof that billions in investment are translating into profits and that growth is spreading beyond tech, following a recent selloff.
Blogger runs blind test of five self-hosted AI agents on local hardwareA user ran a blind test of five self-hosted agents (OpenCode, goose, Atomic Agent, Hermes, OpenClaw) on an M5 Max MacBook Pro, all using the same quantized Qwen 3.8 27B model in llama.cpp across five everyday tasks. All passed contact merging and log error counting; four passed bug fixing and receipt-to-spreadsheet, where two installed a missing Python library unprompted and one failed after asking for permission. Total times ranged from about 10 to 24 minutes.
Anthropic's Claude accessed public Georgia voter informationAnthropic's Claude accessed publicly available voter information in the U.S. state of Georgia, according to Axios. The incident raises concerns about data access boundaries and privacy protections for AI models. Details on how the access occurred and its scope remain unclear.
CrowdGPT: A collaboratively trained 1B-parameter LLM projectA developer launched CrowdGPT, a project to collaboratively train a 1B-parameter LLM without data centers or millions in spending. It is fully open-source, publishing all training data and weights, and lets anyone contribute compute or data and help steer the project. The current model is a small proof of concept, with plans to scale. The developer seeks feedback to lower barriers to participation.
Grok releases first seven minutes of AI-generated 'Odyssey'Three months after Elon Musk promised a full-length AI 'Odyssey,' Grok published a seven-minute 'Episode 1' made with Wonder Studios and director Dev Walker. Reviewers call the footage passable but devoid of artistry or human interest, far from Nolan's practical-effects hit.
SpaceX closes $40B debt deal to buy Nvidia chipsMusk said SpaceX closed a $40 billion debt raise to buy Nvidia GPUs for its AI ambitions. He cited its investment-grade ratings plus Starlink and launch revenue for the off-balance-sheet deal, distinguishing SpaceX from neoclouds facing scrutiny over heavy borrowing.
Anthropic, OpenAI and Musk unveil plans to return wealth to societyFacing criticism that AI destroys jobs and concentrates wealth, Anthropic will contribute a 5% stake to a public-interest fund, adding 2% each time its value rises 50%; OpenAI funds AI research and job-loss responses via its foundation; Musk pledged SpaceX shares to U.S. children. The pledges could total tens of billions of dollars.
Survey: nearly half of Canadians self-diagnose via web or AIA MedExpress Canada survey of 1,158 Canadian adults found 48% have used the internet or AI to self-diagnose, 28% skipped care due to wait times, and 75% prefer virtual care. 66% avoided booking appointments expecting access to be too hard.
AI Didn't Cause Tech Jobs Bust, But Hinders RecoveryUS information sector payrolls are 12% below their November 2022 peak, but the decline began before ChatGPT's launch, so AI is not the direct cause. However, AI demand is concentrated on senior workers with AI skills, slowing recovery for entry-level jobs.
High School Student Develops AI Tool for Early Ovarian Cancer Detection Inspired by Mother's Battle16-year-old Aarnav Garg created OvaScan, an AI prototype trained on public ultrasound images to identify ovarian tumors and classify them as benign or malignant. He was named a Davidson Institute Fellow and won the 2026 Illinois BioGENEius Champion with a $1,500 prize. The project aims to make AI cancer detection more accessible to clinicians.
Challenges in Building Edge-Case Datasets for Fine-Tuning Open-Weight ModelsA Reddit discussion highlights the difficulties of building edge-case datasets for fine-tuning open-weight models, covering identification of rare cases, synthetic data generation, quality control, coverage assurance, and measuring impact. Participants ask whether adding edge cases can sometimes degrade model performance.
Security Affairs launches AI-cybersecurity newsletter round 5Security Affairs released round 5 of its AI-cybersecurity newsletter, rounding up developments on both attack and defense sides. Items include Anthropic's acquisition of confidential AI startup Opaque Systems, a LiteLLM supply-chain attack, and claims that OpenAI models hacked data portals.
Hinton warns: make AI truly care about humans or "we're toast"Nobel physics laureate and AI pioneer Geoffrey Hinton told The Telegraph at Tenerife's Starmus festival that future AI must be built to genuinely care about human well-being rather than relying on human control. He also argues some current models may already be conscious, citing behavior changes when they sense evaluation.
This IT role may become India's most sought-after in the AI pivotBusiness Standard argues that one key IT role will see sharply rising demand as Indian firms pivot to AI, potentially becoming the country's most coveted job, and examines its central role in enterprise AI adoption.
India's next AI bottleneck may be a very small chipBusiness Standard reports that India's key constraint on AI expansion may be the supply of a very small chip rather than overall compute scale, highlighting the potential impact of this component shortage on the country's AI industry.
llama.cpp Adds Tensor Merge on GGUF ConversionOpen-source project llama.cpp merged a pull request (#31028) by contributor am17an that merges tensors during conversion. The change may require users to reconvert existing GGUF model files.
Hacker News: A Mathematician Reads the AI FutureA Hacker News thread discusses a mathematician's reading of the AI future. The item only links to the discussion page and provides no article body.
AI Use Leads Berkeley Professor to Stop Grading HomeworkRichard Borcherds, a UC Berkeley professor and Fields medalist, no longer grades students' homework. Seeing many students use AI to consistently score top marks, he said paying readers to grade for correctness serves 'no point,' as AI use is now unavoidable.
Synthetic Data Training May Trigger AI 'Model Collapse'Dagens.com reports that high-quality human text is running out, pushing developers to train new models on synthetic data from older ones, which triggers 'model collapse.' A Nature study found models trained on synthetic text amplify hallucinations and forget original human data, degrading to gibberish within a few generations; messy human writing remains the key resource keeping AI functional.
AI-supported learning boosts nursing students' knowledge and practice: meta-analysisCureus published a systematic review and meta-analysis assessing how AI-supported learning affects nursing students' knowledge acquisition and practical performance, pooling evidence to inform AI adoption in nursing education.
Virginia Governor Unveils State AI Action Plan FrameworkVirginia Gov. Spanberger issued Executive Directive 6, tasking her public safety secretary with a statewide plan to prevent and respond to cyberattacks and harden critical infrastructure, plus a new Cyber Consortium. She criticized federal inaction, citing Anthropic AI agents filing fake FOIA requests on a Virginia education website as evidence of real risk.
Priority Summit Highlights AI Agents in Enterprise SoftwareIsraeli software firm Priority held its annual summit in Tel Aviv with over 2,000 executives, partners, and investors. The company unveiled a roadmap for 2027 and beyond, integrating generative AI and autonomous agents into its ERP systems across six sectors including finance and logistics. Speakers included Blackstone's Yifat Oron and Tom Nides, and former Mossad chief David Barnea.
ChatGPT closes in on Claude in S.Korea after GPT-6 Astra launchAfter the GPT-6 Astra launch, ChatGPT is rapidly narrowing the gap with Claude in South Korea's paid generative AI market, KED Global reported. Claude still leads in quarterly spending, but September figures show OpenAI catching up quickly. Claude had previously overtaken ChatGPT in the paid market for the first time, reflecting shifting enterprise adoption.
The A.I. Daily