AIskimIQ

Denní přehled zpráv z AI & technologií

Archiv přehledů/čtvrtek 2. července 2026

čtvrtek 2. července 2026

41 článků

Shrnutí událostí za posledních 24 hodin

Přehled zpráv z oblasti AI a technologií za 2026-07-02

Poslechněte si brief jako podcast
Martin Ševčík

Připravil Martin Ševčík
2. července 2026 v 09:41

Poslední dny vidím zajímavý trend, kterému se dosud moc pozornosti nedostalo: velkých firem si začínají stavět vlastní jazykové modely místo toho, aby se spoléhaly na obecné řešení od OpenAI nebo Anthropicu. A myslím, že to není jen výzvou pro etablované AI labory, ale znamená to také zásadní změnu v tom, jak se budou budovat aplikace.

Travelers to udělali s pojišťovactwím. Vyvinuli TravelersLLM, model přímo vyladěný na jejich obchodní procesy, data o pojištění a specifickou terminologii. To dává smysl – pojišťovna má miliony stran interních dokumentů, desetiletí znalostí o underwritingu, procesy, které se neliší od těch, co dělá konkurence pouze tím, že jsou různě balené. Proč by jejich majitelé byli spokojení s modelem trénovaným na obecném internetu, když si mohou koupit lepší nástroj? Apropos, Base44 jde stejným směrem – jejich vibe coding startup si postavil vlastní model, protože zjistili, že obecné modely vyprodukují jen sadu klišé designů. Stavějí vlastní, aby jejich uživatelé mohli psát weby, které skutečně vypadají jinak.

To, co mě na tom zajímá, není samotná technologie. Je to implikace: pokud každá větší organizace bude mít vyladěný model na míru svému byznysu, jak to změní konkurenci? Nebudou už všichni postaveni na stejném základě. Startup, který si nemůže koupit své školení, bude znevýhodněn. A zároveň to znamená, že generalistické modely mohou být víc pro širší masu – pro texty, kód, analýzu – zatímco specialisté jdou jinam.

Evropský rozměr téhož problému? EU si teď chce vytvořit vlastní open-source model pokrývající čtyřiadvacet jazyků přes EUROPA konsorcium. Chápu motivaci – nechcete být závislí na Silicon Valley infrastruktuře pro základní lingvistické schopnosti vašeho kontinentu. Ale otázka zní, zda open-source projekt zvládne konkurovat tlaku vlastních, uzavřených modelů, které si budují jednotlivé státy nebo korporace. Hybrid approach s místní privatností a cloudovým uvažováním je zajímavý, ale praktická dominance zůstane u těch, kdo mají data a rozpočet na trénink.

Antropic zase pustil Claude Sonnet 5 s vylepšeními v kódování a bezpečnosti. To je inkrementální, solidní pokrok. Zajímavě už nejsou tak přísní s kontamináciou stylů – zvedli omezení Fable a Mythos. Znamená to, že si věří v kvalitu baseline modelu a chce jim dát vývojářům víc volnosti. Rozumím tomu, ale přeji si, aby byli transparentnější o tom, co se tím změní v praxi.

Jádro věci: běžná LLM se standardizuje a specifikuje. Průmysl si bere kontrolu do vlastních rukou.

Seznam odkazů na zdroje použitých v přehledu

Zpráva

Revvity Signals brings AI-powered R&D data access to Anthropic's Claude

Revvity has expanded its AI strategy by bringing its Revvity Signals Software business into Anthropic's Model Context Protocol (MCP) connector directory. indianchemicalnews.com

Zpráva

Portugal debuts first open-source AI model as Europe pushes for tech sovereignty

One of the European country has launched its first open-source artificial intelligence model to limit the U.S. reliance in sovereignty push. thenews.com.pk

Zpráva

Pharmalittle: We're reading about Anthropic drug development goals, Trump drug plan snags, and more

Belgium approved a new medicine to treat kidney disease through its Early and Equitable Fast Access procedure for the first time. statnews.com

Zpráva

EU Backs Open-Source AI Model Covering 24 EU Languages

The European Commission has selected the EUROPA consortium, led by Italian AI company Domyn, to develop an open-source large language model (LLM) covering... slator.com

Zpráva

Anthropic launches Claude Sonnet 5 AI model with coding, safety upgrades as Fable and Mythos controls lifted

Anthropic PBC today debuted Claude Sonnet 5, a midrange large language model that outperforms its predecessor in several areas. The LLM will be the default... siliconangle.com

Zpráva

Base44's CEO says the company built its own model to stop churning out AI-slop designs

Base44 wants to help users ditch the cookie-cutter look of AI-coded websites with its new AI model. The San Francisco-based vibe coding startup announced on... businessinsider.com

Zpráva

Travelers Builds Insurance-Specific LLM

The Travelers Companies, Inc. announced this week that engineers and data scientists have developed TravelersLLM, a proprietary large language model... carriermanagement.com

Zpráva

Travelers Advances AI Strategy with Award-Winning Insurance-Specific Large Language Model

The Travelers Companies, Inc. (NYSE: TRV) today announced that it has developed TravelersLLM, a proprietary large language model tailored to its property... businesswire.com

Zpráva

Hybrid LLM Workflows Blend Local Privacy with Cloud Reasoning

A detailed field guide published on Towards Data Science explores hybrid local-cloud large language model (LLM) workflows, offering a practical framework... techgig.com

Zpráva

Travelers unveils proprietary insurance-focused large language model

Travelers has developed TravelersLLM, a proprietary large language model designed specifically for its property and casualty insurance business. coverager.com

Zpráva

Why we built ADK 2.0

Moving AI Agents from prototype to production creates new challenges. In real-world enterprise environments, agents can get stuck in infinite loops,... developers.googleblog.com

Zpráva

Godot has implemented a ban on contributions involving "autonomous AI agent use or vibe coded" elements

Godot has revised its contribution policy and has implemented measures to curb "demoralizing" AI contributions. wnhub.io

Zpráva

Godot bans "autonomous AI agent use or vibe coded" contributions

Godot has updated its contribution policy and "take[n] steps" to reduce "demoralizing" AI contributions. gamesindustry.biz

Zpráva

Mastering Agentic Techniques: AI Agent Reinforcement Learning | NVIDIA Technical Blog

Reinforcement learning (RL) is central to aligning language models, from reinforcement learning with human feedback (RLHF) within AI assistants to newer… developer.nvidia.com

Zpráva

Pensacola launching AI agent 'Penny' to field 311 calls

City Administrator David Stafford said the AI agent is the first of other AI initiatives the city is considering. pnj.com

Zpráva

Evaluating Agentic AI Across the Lifecycle on OCI

Agentic AI systems do more than generate answers: they retrieve evidence, call tools, use identity, change state, and recover from failures. blogs.oracle.com

Zpráva

Secure Agentic Access: Authentication and Authorization for AI Agent Workloads

8 min readMany organizations deploying AI agents have not fully answered how those agents authenticate to services, what credentials they carry,... securityboulevard.com

Zpráva

Shilo rolls out autonomous 1:1 AI coaching for agent calls

Shilo launched 1:1 Coaching, an autonomous voice-to-voice AI sales coach for real estate and mortgage teams. The tool uses each agent's call data and DISC... housingwire.com

Zpráva

AI-Generated Presentations: Can Better Prompts Save Time?

Structured prompts help physicians prepare conference presentations faster, but expert review remains essential to verify evidence, references,... medscape.com

Zpráva

Four-Step Test Detects AI Errors Before Strategy

Search Engine Journal contributor **Alexander Kesler** published a four-step verification protocol on **July 1, 2026** designed to catch AI errors,... letsdatascience.com

Zpráva

New AI Agents Pose ‘Existential Threat’ to How Grants Are Awarded

The rise of more sophisticated artificial intelligence agents poses an “existential threat” to the way research funding is awarded, experts have warned,... insidehighered.com

Zpráva

[Sci-Tech NOW] POSTECH develops methodology to measure LLM uncertainty

From left: Professor Namhoon Lee of the POSTECH Graduate School of Artificial Intelligence, integrated PhD student Jinseok Jeong, and master's students... dongascience.com

Zpráva

Jumpmind CISO Launches "CIRCUIT" Open-Source AI Framework

Jumpmind's CIRCUIT is an open-source framework that audits autonomous AI decisions and turn "black box" algorithms into inspectable security controls. opensourceforu.com

Zpráva

Reinstating death penalty is an "existential risk" to democracy, warns Macron

President Emmanuel Macron warns that the death penalty is not a deterrent and labels its potential return an existential risk to democracy. plataformamedia.com

Zpráva

AI finds hidden ECG signal that predicts sudden cardiac death risk

Sudden cardiac death kills more than 300000 people in the U.S. each year, even though implantable defibrillators have been able to stop many lethal... health.yahoo.com

Zpráva

Sandeep Reddy outlines his approach to developing real-world AI applications

Most people experience artificial intelligence at the surface level, quick responses, smooth recommendations, systems that seem to “just work. tribuneindia.com

Zpráva

Set AI credit session limits in Copilot CLI and SDK

You can now set AI credit session limits in Copilot CLI and the GitHub Copilot SDK to cap the amount an agent spends in a session. This is especially useful... github.blog

Zpráva

Sopra Steria Elevates Service Center Performance with NiCE, Advancing AI-Driven Operational Excellence Across Europe

With NiCE CXone and Copilot, Sopra Steria has a secure platform which consolidates communication channels into a single agent interface where 90% of calls... businesswire.com

Zpráva

TripGain Launches AI Spend Copilot to Help Enterprises Turn Expense Data into Real-Time Decisions

PRNewswire/ -- TripGain, a fast-growing AI-powered travel and expense management platform, recently launched AI Spend Copilot, a conversational intelligence... prnewswire.com

Zpráva

Agentic Resource Discovery

Agentic resource discovery - Agentic resource discovery is reshaping how developers interact with AI assistants by enabling GitHub Copilot to automatically... trendhunter.com

Zpráva

Tech Moves: Amazon Music names VP; Microsoft departures and a Copilot shakeup; Veeam adds exec

Amazon Music names VP of product and tech; Microsoft's corporate VP of Security, Compliance, Identity, Management & Privacy steps down while its Copilot... geekwire.com

Zpráva

Attentive.ai launches BIM CoPilot, bringing human-vetted BIM management to the build phase

San Francisco, CA: Attentive.ai today announced the launch of BIM CoPilot. It is a human-vetted, multi-trade BIM management solution that. geoweeknews.com

Zpráva

Microsoft Brings Service Agent GA to Microsoft 365 Copilot

The GA release unifies Dynamics 365 and Microsoft 365 workflows, giving service reps AI-driven access to case data and actions in one interface. cmswire.com

Zpráva

Euromonitor revamps Passport with AI chat & Copilot

Euromonitor has launched an AI-based version of its Passport market intelligence platform, overhauled with integrations for Microsoft Copilot and other AI... itbrief.asia

Zpráva

Logicalis strengthens enterprise AI capability with Global Microsoft Frontier Partner status and Copilot specialisation

Logicalis, the global technology service provider, has announced it has achieved Microsoft Frontier Partner status, alongside its Microsoft Copilot... itwire.com

Zpráva

OpenAI looks beyond a single ad format with image, video and conversational ads in the works

OpenAI is hiring San Francisco-based software engineers to build text, image video, native, conversational and interactive ads. digiday.com

Zpráva

G6 Hospitality launches AI-powered video tours

G6 Hospitality, parent company of Motel 6 and Studio 6 brands, has launched Vision 360, a subscription that delivers video tours of franchise properties. hotelmanagement.net

Zpráva

Google Launches AI Image and Video Tools: 4-Second Images, Conversational Video Editing

Google released two generative media models on Tuesday — Nano Banana 2 Lite for rapid image generation and Gemini Omni Flash for conversational video... techtimes.com

Zpráva

Google introduces a faster, cheaper image generator with Nano Banana 2 Lite

Google on Tuesday released Nano Banana 2 Lite, the newest version of its in-house AI video and image generator. This version is significantly faster and... techcrunch.com

Zpráva

Google launches faster AI image, video tools to give users’ more creative powers

Google has unveiled two new artificial intelligence models to make AI-powered image and video creation faster, cheaper, and easier to access for developers,... businessday.ng

Zpráva

AI Music Video Generation: 10 Tools That Automate Your Creative Workflow in 2026

Creating a music video no longer requires expensive software or advanced editing skills. Now, AI can turn ideas, songs, or images into engaging videos in... roboticsandautomationnews.com

Podpořte projekt

AIskimIQ je nezávislý projekt. Pokud vám přijde užitečný, můžete jeho rozvoj podpořit kávou.

Kupte mi kávu ☕