Ten years in the past, elevating €1 million in Copenhagen was sufficient to make waves within the area’s tech scene. Right this moment, the Nordics are turning out billion-dollar corporations like Lovable — which hit $200M in income simply 12 months after launching. Dennis Inexperienced-Lieber, founding father of AI-powered buyer intelligence platform Propane, has had a front-row seat to that shift over the past 15 years. His take? The area’s social security internet provides founders room to take actual swings with out placing their private lives on the road, and they’re accelerating sooner than Silicon Valley because of this. Right this moment on TechCrunch’s Fairness podcast, Dominic-Madori Davis…
Author: Naveed Ahmad
Tencent Hunyuan has launched HunyuanOCR, a 1B parameter imaginative and prescient language mannequin that’s specialised for OCR and doc understanding. The mannequin is constructed on Hunyuan’s native multimodal structure and runs recognizing, parsing, info extraction, visible query answering, and textual content picture translation via a single finish to finish pipeline. HunyuanOCR is a light-weight different to common VLMs corresponding to Gemini 2.5 and Qwen3 VL that also matches or surpasses them on OCR centric duties. It targets manufacturing use circumstances like doc parsing, card and receipt extraction, video subtitle extraction, and multilingual doc translation. https://github.com/Tencent-Hunyuan/HunyuanOCR/blob/essential/HunyuanOCR_Technical_Report.pdf Structure, Native Decision ViT plus…
Over 1,000 Amazon staff have anonymously signed an open letter warning that the corporate’s allegedly “all-costs-justified, warp-speed strategy to AI improvement” may trigger “staggering harm to democracy, to our jobs, and to the earth,” an inside advocacy group introduced on Wednesday.4 members of Amazon Workers for Local weather Justice inform WIRED that they started asking staff to signal the letter final month. After reaching their preliminary purpose, the group revealed on Wednesday the job titles of the Amazon staff who signed and disclosed that greater than 2,400 supporters from different organizations, together with Google and Apple, have additionally joined in.Backers…
The software program group at Common Motors has now misplaced three prime executives up to now month because the automaker — with its new chief product officer on the helm — combines its disparate know-how companies into one group. Baris Cetinok, senior vice chairman of software program and providers product administration, is leaving the corporate efficient Dec. 12, the corporate confirmed to TechCrunch. CNBC was the first to report the departure. Dave Richardson, senior vice chairman of software program and providers engineering, and Barak Turovsky, who was employed in March as head of AI on the automaker, have additionally left…
Synthetic intelligence startup xAI, based by Elon Musk, instructed metropolis and county planners in Memphis final week that it plans to construct a photo voltaic farm subsequent to its Colossus information middle — one of many world’s largest services for coaching AI fashions. The undertaking would occupy 88 acres to the west and to the south of the information middle. A 136-acre vacant lot owned by the developer that additionally owns the Colossus property at the moment borders the location. Given the proposed measurement, the photo voltaic farm would probably produce round 30 megawatts of electrical energy, solely about 10%…
A number of public web sites designed to permit courts throughout the US and Canada to handle the non-public info of potential jurors had a easy safety flaw that simply uncovered their delicate knowledge, together with names and residential addresses, TechCrunch has completely realized. A safety researcher, who requested to not be named for this story, contacted TechCrunch with particulars of the easy-to-exploit vulnerability, and recognized a minimum of a dozen juror web sites made by authorities software program maker Tyler Applied sciences that seem like susceptible, provided that they run on the identical platform. The websites are all around…
In response to questions on its two-phase cooling merchandise from WIRED, together with whether or not or not the corporate deliberate to submit chemical compounds for fast-tracked consideration below the administration’s new information middle exemption, Chemours spokesperson Cassie Olszewski mentioned the corporate is “within the technique of commercializing our two-phase immersion cooling fluid, which would require related regulatory approvals.”“Our work on this space has been centered on creating extra sustainable and environment friendly cooling options that might enable information facilities to devour much less vitality, water, and footprint whereas successfully managing the growing quantity of warmth generated by the subsequent…
Main tech corporations aren’t simply utilizing AI that will help you generate or summarize content material — additionally they need you to make use of it for purchasing. OpenAI, Google, and Amazon have closely invested in AI assistants that analysis new product classes for you and counsel the suitable ones to purchase. Startups like Perplexity, Daydream, and Cherry have additionally constructed companies round AI for product discovery. All these efforts have resulted in clients utilizing extra AI for purchasing. Onton (beforehand often called Deft), an AI-powered furnishings purchasing platform, says it has seen its person base develop from 50,000 month-to-month…
When your utility can name many alternative LLMs with very completely different costs and capabilities, who ought to resolve which one solutions every request? Salesforce AI analysis crew introduces ‘xRouter’, a tool-calling–based mostly routing system that targets this hole with a reinforcement studying based mostly router and learns when to reply regionally and when to name exterior fashions, whereas monitoring price at token degree. What’s xRouter? xRouter is a instrument calling based mostly orchestration system constructed on Qwen2.5-7B-Instruct because the router spine. The router is an instruction tuned mannequin with instrument calling capabilities that decides which downstream mannequin to invoke,…
On this tutorial, we discover find out how to construct neural networks from scratch utilizing Tinygrad whereas remaining totally hands-on with tensors, autograd, consideration mechanisms, and transformer architectures. We progressively construct each part ourselves, from primary tensor operations to multi-head consideration, transformer blocks, and, lastly, a working mini-GPT mannequin. Via every stage, we observe how Tinygrad’s simplicity helps us perceive what occurs below the hood when fashions prepare, optimize, and fuse kernels for efficiency. Try the FULL CODES here. import subprocess, sys, os print(“Putting in dependencies…”) subprocess.check_call([“apt-get”, “install”, “-qq”, “clang”], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) subprocess.check_call([sys.executable, “-m”, “pip”, “install”, “-q”, “git+https://github.com/tinygrad/tinygrad.git”]) import numpy as…