AI agents fail to meet 2025 workforce predictions

OpenAI CEO Sam Altman predicted in early 2025 that AI agents would join the workforce and materially change company output, but the technology has fallen dramatically short by year's end according to industry insiders.

Andrej Karpathy, an OpenAI co-founder, said in October that AI agents are "cognitively lacking and it's just not working," estimating it will take about a decade to resolve current issues with intelligence, multimodal capabilities, and memory retention.

A Deloitte survey found only 11% of organizations currently use AI agents in production, with real-world testing revealing even basic tasks like clicking elements or selecting dropdown menus can take minutes or fail entirely.

Read more

Reasoning models and coding agents reshape AI in 2025

Reasoning models that employ multi-step logical analysis before generating responses became the defining AI technology of 2025, with DeepSeek's R1 model released in January demonstrating how to build such capabilities and triggering rapid global adoption.

Autonomous coding agents evolved to handle complex software development tasks, enabling Amazon to save approximately 4,500 developer-years of work and $260 million annually while models like Claude Opus 4.5 completed more than 80 percent of coding challenges by year's end.

Data center construction and AI infrastructure investments accounted for up to 92 percent of U.S. GDP growth in the first half of 2025, with Microsoft committing $80 billion and Meta spending approximately $72 billion on global data center projects.

Read detailed article

Z.ai releases open-source coding model topping benchmarks

Z.ai released GLM-4.7, an open-source large language model, designed for real-world software development with enhanced code generation, multi-step reasoning, and tool-calling capabilities.

The model scored 87.4 on τ²-Bench and ranked first among open-source models in Code Arena's blind evaluation of over one million comparisons, with overall performance approaching Claude Sonnet 4.5.

The Beijing-based startup, which passed a Hong Kong Stock Exchange listing hearing in mid-December and is expected to raise $300 million, affirmed its commitment to open-source development despite going public.

Read more

Google fires Korea executives after failing to secure AI chip supplies

Google fired senior procurement executives in Korea after they failed to secure long-term supply agreements for high-bandwidth memory chips, with Microsoft and Meta also stationing purchasing teams in South Korea amid an intensifying AI chip shortage.

Only three companies worldwide—SK Hynix, Samsung Electronics, and Micron —can produce advanced HBM and high-performance DRAM, and both Korean manufacturers have sold out their entire production capacity through 2026.

The crisis has prompted tech giants to relocate procurement roles from Silicon Valley to Asian manufacturing hubs and place open-ended orders "regardless of price," with industry sources warning that advanced production lines are running at full capacity and cannot meet all demands.

Read more

😷 Up to 400 Million Face Long-COVID Brain Symptoms

A review published Monday in Nature Reviews Disease Primers highlights the massive global impact of long COVID on the brain. It estimates that 80 to 400 million people worldwide suffer from lasting neurological issues months or years after infection.

The study notes that symptoms often continue for at least three months. Common problems include memory loss, anxiety, depression, headaches, and sleep disturbances. The condition affects 5–20% of general cases and up to 50% of hospitalized patients.

Have you experienced Long-COVID yourself?


Read More

#CryptoSecurity #TrustWallet #BrowserExtension
Trust Wallet Chrome extension hack tied to millions in losses

A compromised Trust Wallet Chrome extension update on December 24 drained wallets and fueled a parallel phishing spree.

Key highlights
- Extension version 2.68 was abused before a rapid fix to 2.69
- Estimated crypto losses run into millions
- Hidden code exfiltrated wallet data to an external endpoint
- Phishing sites (e.g., fix-trustwallet.com) targeted victims for seed phrases

Read More: Trust Wallet Chrome extension hack tied to millions in losses

BleepingComputer

Trust Wallet confirms extension hack led to $7 million crypto theft

Several users of the Trust Wallet Chrome extension report having their cryptocurrency wallets drained after installing a compromised extension update released on December 24, prompting an urgent response from the company and warnings to affected users. Simultaneously…

Google's Gemini triples market share as ChatGPT's lead shrinks

Google's Gemini more than tripled its share of generative AI web traffic to 18.2% over the past year while ChatGPT's dominance plummeted from 87.2% to 68%, according to analytics firm Similarweb.

The 19-point market share loss prompted OpenAI CEO Sam Altman to declare a "Code Red," fast-tracking the December 11 release of GPT-5.2 and delaying other initiatives like advertising and AI agents.

Analysts attribute Gemini's surge to its integration across Chrome, Android, Google Workspace, and Search, placing the AI directly in users' existing workflows rather than requiring a separate app, while Alphabet stock climbed over 60% year-to-date.

Read more

X lets users edit anyone's images with AI, sparking artist backlash

X rolled out a feature on December 24, 2025, allowing any user to AI-edit images posted by others using its Grok chatbot, sparking outrage from artists who say there is no opt-out mechanism.

The feature lets users long-press public images to prompt edits that are posted as comments, with artist protection tools like Glaze and Nightshade proving ineffective against the AI alterations.

Artists including manga creator Boichi announced plans to stop posting work or migrate to rival platform Bluesky, while Elon Musk promoted the feature amid backlash, posting "Try Grok image edit."

Read more

OpenAI plans ads in ChatGPT responses, reversing earlier stance

- OpenAI is developing ad formats for ChatGPT that could embed sponsored content directly into AI responses or display ads in sidebars, with a planned 2026 rollout targeting its 800 million weekly users.​

- The move reverses CEO Sam Altman's previous stance calling AI-powered advertising "dystopian," as the company faces mounting costs with $2.5 billion burned in early 2025 and infrastructure spending projected to exceed $17 billion in 2026.​

- Internal mockups show ads appearing alongside responses with clear disclosures, while the company explores using ChatGPT's memory function for personalized targeting based on users' conversation histories.

Read more

❤️ World Record for Longest Remote Robotic Surgery Set

A surgeon in Kuwait successfully operated on a patient in Brazil, connecting two rooms across 12,035 km in near real‑time.

This achievement sets a new world record for the most significant distance between a doctor and a patient during a robotic operation.

Read More

✈️ Durov Offers Free IVF Using His Sperm

Telegram founder Pavel Durov announced he will pay for in vitro fertilization (IVF) for unmarried women aged 37 or younger who use his sperm. The 41‑year‑old billionaire, who has already fathered over 100 children, calls the offer a way to help solve the global fertility crisis.

The program is managed by Moscow's Altravita Clinic, which promotes it as a unique chance to use donor material from a famous entrepreneur. Durov is covering the full cost of the treatments, which usually run around $5,000.

What do you think?

❤️ — Marketing
🔥 — He's just... being Pavel!

Read More

☃️ Disney's AI Olaf Robot Now Greets Guests in Paris

The AI robot Olaf from Frozen now greets guests at Disneyland Paris. Its smooth, realistic movements are the result of reinforcement learning and Nvidia technologies.

The robot is trained in a simulation, where it learns to accurately reproduce movements crafted by animators.

Read More