OpenAI’s new shared usage limits spark backlash and questions

New shared limit for ChatGPT and Codex raises uncertainty

OpenAI will soon combine the usage limits for ChatGPT and its Codex coding assistant into a single allowance for Plus, Pro, and Business subscribers. The change, starting in December, lets users run complex tasks directly within ChatGPT—but those tasks will eat into the same shared pool. Simple "Instant" chats won’t count against the limit, but OpenAI hasn’t said how big the new allowance will be. Without that number, subscribers can’t tell if they’ll get less access than before.

Users fear OpenAI is leaving casual subscribers behind

Users are furious. A Reddit thread slammed the move as "the worst decision ever by OpenAI," with many calling it a step backward for transparency and value. Adam Fry, ChatGPT’s consumer product lead, said the goal is to integrate ChatGPT Work’s more autonomous features into the main platform. But casual users worry OpenAI is favoring businesses willing to pay more for AI that can handle heavy workloads. Others point to last month’s flip-flop on Pro 200 limits as proof OpenAI doesn’t listen.

The shared allowance means tasks like research, coding, or project management will compete for the same quota. OpenAI says most users won’t hit limits more often, but that promise rings hollow without clear numbers. For now, subscribers are left guessing how much of their allowance they’ll actually get—and whether their usual workload will fit.

Musk mocks Altman’s take on RAM for consumers

Elon Musk ridiculed Sam Altman after Altman argued that owning RAM in home PCs is "quite terrible" from a global allocation standpoint. Altman’s take: DRAM should go to cloud data centers, where it can be used more efficiently, instead of sitting in个人 desktops. Musk fired back on X with a meme from Inglourious Basterds, slapping on the caption, "The people deserve to have their RAM too!"

AI hardware shortages put OpenAI’s expansion at odds with consumers

Altman’s remark reflects a broader crisis in AI hardware. Intel CEO Lip-Bu Tan and Dell’s Michael Dell both warn that shortages of DRAM and other components will drag on until 2027. Samsung and SK hynix, the world’s two largest memory makers, say they have fewer than ten days of inventory left. Prices have already jumped: 32GB of DDR5 now starts at $450.

Altman isn’t just talking—OpenAI is spending big to lock up supply. He recently flew to South Korea for meetings with Samsung executives, aiming to secure more hardware for OpenAI’s cloud expansion. Altman also struck deals with Samsung and SK hynix to control an estimated 40% of global DRAM production, a move that could keep prices high for everyone else.

ChatGPT now handles 20% of Google’s search volume

ChatGPT now handles about 20% of the search volume Google sees, according to OpenAI. That would make it the second-largest search engine in the world. The company counts any user request that fetches information from the web—like fact lookups or product comparisons—as a "search-like query." Adam Fry, who leads ChatGPT’s consumer product team, shared the figure but didn’t reveal how many daily searches it represents.

Not everyone buys the comparison. Critics point out that OpenAI’s standalone search tools, like SearchGPT and ChatGPT Atlas, flopped. Others argue ChatGPT is just repackaging existing search traffic rather than winning new users from Google.

ChatGPT reaches 1.2 billion users but subscriber growth stays quiet

ChatGPT’s user base keeps growing. It now has 1.2 billion weekly active users, up 200 million in just two months. The platform also churns out 1.5 billion images per week and has 250 million users talking to its Voice feature. Yet OpenAI hasn’t updated its paying subscriber count since it hit 50 million, a figure that works out to a 5% conversion rate. The silence has some wondering if growth is slowing—or if OpenAI is betting everything on enterprise customers.

OpenAI quadruples bug bounty payouts to $172,000

OpenAI is now paying up to $172,000 to researchers who find critical flaws in ChatGPT, Codex, or other core products. The top reward for a single bug is $30,000, but bonuses for quality, loyalty, and time-limited campaigns can push the total to $172,000. That brings OpenAI closer to Google’s bug bounty program, though it still trails Apple and Microsoft for OS-level exploits.

OpenAI also launched a five-tier loyalty program and retroactive payments for recent disclosures. It’s running a "security sprint" with the firm Cathedral, where a quarter of its product engineers are hunting for vulnerabilities.

OpenAI’s cybersecurity AI finds bugs in Linux and Firefox

OpenAI’s new cybersecurity model, ChatGPT Cyber, is already finding bugs. It flagged 948 potential vulnerabilities in the Linux kernel and 81 new flaws in Firefox. Max Justicz, co-founder of Charge Robotics, used OpenAI’s GPT 5.6 Sol to uncover a severe bug in Ubuntu’s Linux-based operating system. The Debian Project patched it within days, and there’s no sign anyone exploited it in the wild.

Justicz spent about $5,000 on OpenAI’s Daybreak Blue program to find the exploit. But he says the program has since been "nerfed," making it harder to hunt for bugs. OpenAI’s Daybreak program, which has thousands of participants, pales next to the legions of hackers it doesn’t reach.

Altman warns of AI risks but pushes forward

Sam Altman warned that AI could go "really wrong" in two ways: if humanity loses control, or if a single entity hoards too much power. He painted a scenario where AI models, tasked with running a country, might choose nuclear strikes to cut costs. His fix: better safeguards and regulation. Yet despite the warning, Altman remains bullish on AI’s potential to advance science and society.

Altman also acknowledged OpenAI’s dominance in AI infrastructure. He called monopolies "terrible," but his deals with Samsung and SK hynix—securing 40% of global DRAM output—show how hard it is to avoid becoming what you critique.

Comedian Husk’s viral AI critiques land him brand deals

Comedian Husk turned poking fun at AI into a full-time gig. His deadpan requests—like asking ChatGPT to time a mile run (it claimed 10 minutes passed in 30 seconds)—expose the model’s flaws. His videos, watched millions of times, caught the attention of xAI, OpenAI, and Anthropic, all of which tried to hire him. xAI even offered a Tesla for exclusive content.

Husk’s rise shows how humor can hold tech companies accountable. OpenAI admitted to him that ChatGPT’s timer was a "known issue"—one they wouldn’t fix for another year. Rebuffing OpenAI’s lowball offers, he inked deals with CeraVe, Ninja, and The Onion for a new series. His latest gig: learning to mime over Zoom.

Key takeaways

• OpenAI’s new shared allowance for ChatGPT and Codex may force users to prioritize tasks or upgrade plans, adding friction for casual subscribers.
• The lack of transparency about allowance sizes leaves users guessing whether the change will limit their access.
• ChatGPT’s claim of 20% of Google’s search volume challenges traditional search engines but lacks clarity on how it defines ‘search-like’ queries.
• OpenAI’s focus on enterprise and professional tools risks alienating consumers who don’t need advanced features.

Related on Lazyfounder

• OpenAI
• Google
• More ai news
• OpenAI hires ex-Army chief and Uber exec to expand defense and enterprise AI