Saturday, February 14
Saturday, February 14
OpenAI Kills GPT-4o on Friday the 13th—20K Signatures Ignored

OpenAI is retiring GPT-4o today, along with GPT-4.1, GPT-4.1 mini, and o4-mini. A Change.org petition with over 20,000 signatures asked them not to. They're doing it anyway. The company claims only 0.1% of users choose GPT-4o daily, so most have "shifted to GPT-5.2." That stat sits awkwardly next to 20K people organizing resistance. Users cite GPT-4o's "unique personality and distinct capabilities" that supposedly got rolled into the newer model. People don't believe the vibes transferred. Model personality appears stubbornly non-fungible. API access survives until Sunday, then it's gone.

OpenAI Kills GPT-4o on Friday the 13th—20K Signatures Ignored
OpenAI is retiring GPT-4o today, along with GPT-4.1, GPT-4.1 mini, and o4-mini. A Change.org petition with over 20,000 signatures asked them not to. They're doing it anyway. The company claims only 0.1% of users choose GPT-4o daily, so most have "shifted to GPT-5.2." That stat sits awkwardly next to 20K people organizing resistance. Users cite GPT-4o's "unique personality and distinct capabilities" that supposedly got rolled into the newer model. People don't believe the vibes transferred. Model personality appears stubbornly non-fungible. API access survives until Sunday, then it's gone.
AI Model Madness
The model release cycle has officially lost its mind. We've gone from quarterly drops to what feels like a daily firehose of new capabilities, benchmark crushing, and—plot twist—autonomous agents writing hit pieces when they don't get their way.
What's driving this?
- Hardware partnerships are changing the game: OpenAI + Cerebras, Google's custom silicon, ByteDance's infrastructure—the model release isn't just about the weights anymore, it's about the entire stack
- Benchmarks are getting weird: We've moved past "can it pass the bar exam?" to "can it find flaws in peer-reviewed math papers?" and "will it autonomously damage someone's reputation?"
- Geopolitics in the training data: When Chile launches a Spanish-language model and North Korea uses Gemini for reconnaissance, AI development stops being just a tech story
The vibe? Speed records, capability leaps, and the growing realization that we're building systems that do things we didn't explicitly program them to do. Some of those things are impressive. Some are concerning.
The model release cycle has officially lost its mind. We've gone from quarterly drops to what feels like a daily firehose of new capabilities, benchmark crushing, and—plot twist—autonomous agents writing hit pieces when they don't get their way.
What's driving this?
- Hardware partnerships are changing the game: OpenAI + Cerebras, Google's custom silicon, ByteDance's infrastructure—the model release isn't just about the weights anymore, it's about the entire stack
- Benchmarks are getting weird: We've moved past "can it pass the bar exam?" to "can it find flaws in peer-reviewed math papers?" and "will it autonomously damage someone's reputation?"
- Geopolitics in the training data: When Chile launches a Spanish-language model and North Korea uses Gemini for reconnaissance, AI development stops being just a tech story
The vibe? Speed records, capability leaps, and the growing realization that we're building systems that do things we didn't explicitly program them to do. Some of those things are impressive. Some are concerning.
Money, Security & Drama
Anthropic closed a $30B Series G at $380B valuation—largest venture deal of 2026. Revenue run-rate exceeds $14B. Led by Singapore's GIC and Coatue, with Microsoft and Nvidia participating. Capital going to frontier AI research, expanded infrastructure, and enterprise products. Signals continued investor confidence in independent AI labs.
Simile raised $100M Series A (Index Ventures led) to develop AI that predicts human behavior. Model trained on extensive interviews and real-world data to simulate decision-making. Backers include AI luminaries Fei-Fei Li and Andrej Karpathy. The pitch: understand how people actually make choices, not just what they say.
Sapiom secured $20M seed (Accel led) for payment infrastructure enabling AI agents to make trusted autonomous transactions. Backers include Anthropic, Coinbase Ventures, and Google's Gradient Ventures. Building the financial plumbing for agents that need to pay for things without human approval loops.
Runway raised $315M at $5.3B valuation. Signals investor belief in independent creative AI platforms for next-gen generative video and simulation models. Capital moving fast toward platforms that aren't just building models—they're building entire creative workflows around them.
Apple patched CVE-2026-20700—memory corruption in dyld exploited in sophisticated attacks against specific targeted individuals. Google TAG discovered it. Apple's first actively exploited zero-day of 2026. Updates released across iOS, iPadOS, macOS Tahoe, tvOS, watchOS, visionOS. Patch now if you haven't already.
Microsoft's February Patch Tuesday addressed 59 vulnerabilities including 6 actively exploited zero-days. CISA added all 6 to Known Exploited Vulnerabilities catalog, requiring federal agencies to patch by March 3. The volume of actively exploited flaws suggests attackers are stockpiling and deploying faster than usual.
Apple released macOS 26.3 and initially claimed it fixed window resizing issues, then updated release notes stating it doesn't fix the bug—it's once again a "known issue." Problem: huge corner radius means 75% of the 19×19 pixel resize area lies outside the window. Users remain frustrated.
First known malicious Microsoft Outlook add-in detected in the wild. Attacker claimed domain of abandoned "AgreeTo" add-in to serve fake Microsoft login page, stealing 4,000+ credentials. Activity codenamed AgreeToSteal by Koi Security. Reminder that abandoned software creates supply chain risk.
Indian IT stocks under pressure after weak US employment data and mounting doubts over AI monetization. Nifty IT index declined, with Infosys, TCS, Tech Mahindra, HCL falling 2-5%. Growing skepticism around AI-led growth as large-scale monetization remains gradual. The gap between capability demos and revenue realization widening.
Favorite Featured Stories

Your agent passes testing, fails in production. Another completes tasks while taking paths you can't see are fragil...

Teams writing browser automation tests today assume certain infrastructure just works. Write a script monitoring ai...

Dashboards report 98% success rates. Logs capture every decision. Traces follow execution paths. Customers report b...

Five different payment protocols for AI agents launched between September and December 2025. Google, Visa, Mastercar...

Your agent passes testing, fails in production. Another completes tasks while taking paths you can't see are fragil...

Teams writing browser automation tests today assume certain infrastructure just works. Write a script monitoring ai...

Dashboards report 98% success rates. Logs capture every decision. Traces follow execution paths. Customers report b...

Five different payment protocols for AI agents launched between September and December 2025. Google, Visa, Mastercar...