If AI can code, why are Big Tech founders coding again?

At first blush, it seems ironic. As artificial intelligence automating coding, some of technology’s most powerful figures are returning to it.

In today’s edition of Mint Tech Talk:

• Hiring slows but data shows AI isn’t killing jobs
• Humanoid beats human record at half-marathon
• AI Tool of the Week: Skills in Chrome
• ChatGPT Images 2.0 ups the stakes

Meta Platforms Inc.’s Mark Zuckerberg, for instance, is working closely with software development teams after two decades. He has moved his desk into Meta’s AI Lab, sitting alongside Alexander Wang and Mat Friedman—co-leads of Superintelligence Labs—and has been actively writing and reviewing code, according to a report by Business Insider. Meta is even building a photorealistic, AI-powered 3D avatar of Zuckerberg to engage with employees in his stead, Financial Times reports.

Google co-founder Sergey Brin is making a similar return. The Information reports that the Gemini maker is assembling an elite “coding strike team” to close the gap with Anthropic, with Brin directly involved. The ambition is to push towards an “AI takeoff” in which systems increasingly improve and code themselves.

Closer home, in January 2025, co-founder Sridhar Vembu stepped down as CEO of Zoho to become chief scientist, focusing once again on writing and reviewing code.

Set against this, Anthropic co-founder and CEO Dario Amodei has argued that coding itself may eventually disappear, replaced by systems that turn human intent directly into working software.

Anthropic is already building toward that future with tools like Claude Code, Mythos (restricted distribution due to security reasons even as OpenAI Inc.’s GPT-5.4-Cyber, released this month, is a model finetuned for additional cyber capabilities and with fewer capability restrictions for those who are part of its Trusted Access for Cyber programme), and Claude Design.

After triggering a selloff in stocks of SaaS companies earlier this year with Claude Code, Anthropic’s new AI tool Claude Design is threatening software design firms like Adobe and Figma. Anthropic says Claude Design can create polished visual assets ranging from interactive prototypes to pitch decks—all with a simple text prompt.

Given these developments, what looks like a contradiction is, in fact, a shift. AI is reshaping how software gets built. Instead of writing detailed instructions, developers increasingly describe outcomes and let machines generate the rest. That makes AI the new interface between humans and computers—a role once held by operating systems and cloud-computing platforms. Whoever controls this interface, in effect, shapes how software is created.

This is precisely why founders are getting closer to software again—not just to write more code but to understand how this new layer behaves.

Consider Meta’s internal targets. According to documents cited by Business Insider, the company expects that by mid-2026, 65% of engineers in its core product groups will generate more than 75% of their code using AI. In a world where small changes in prompts can produce widely different outcomes, rolling up one’s sleeves is not a throwback. It is a strategy.

At the same time, AI is compressing software work. Tasks that once required entire teams can now be handled by a single developer equipped with the right tools. This is not full automation, but a reduction in the number of people needed between idea and execution. Cost too is becoming a constraint: AI-generated code is not free. Tokens carry a price. Every prompt and iteration consumes compute, raising a basic question of efficiency: When is it cheaper to use machines, and when do humans remain more effective?

The answer will shape how quickly AI transforms traditional development. For now, the technology remains imperfect. AI can generate code quickly, but it is still prone to errors, edge cases, and unpredictable behaviour. That makes human oversight more important, not less. The work is shifting from writing code to guiding and validating it.

So what changes? Routine coding tasks will decline. Large engineering teams may shrink. But higher-order skills—system design, judgment, and the ability to translate ideas into precise, executable instructions—will become more valuable. Coding does not disappear; it moves up the stack. The founders returning to it understand this. The future may need fewer programmers, but it will demand more leaders who can think like one.

Hiring slows but data shows AI isn’t killing jobs

The bad news for white-collar workers continues. As per layoff.fyi, a little over 81,000 employees have been laid off by 97 tech firms since the beginning of this year.

LinkedIn too has confirmed a 20% decline in hiring since 2022. But it does not blame AI for the ongoing slowdown, as per a TechCrunch report, citing an interview at the Semafor World Economy Summit. LinkedIn’s Chief Global Affairs and Legal Officer Blake Lawit instead said the broader decline in hiring is more closely tied to rising interest rates.

“At LinkedIn…we have an economic graph which is over a billion members. We’ve got companies, jobs, skills. It’s really an amazing real-time view of what’s happening in the labour market. And we’ve looked—because everyone wants to know the answer to this question: Is AI impacting jobs right now? We’ve looked and, honestly, we haven’t seen it,” Lawit said. He argued that if AI was actively replacing workers, the job losses would be highly visible in specific, vulnerable sectors such as customer support, administrative roles, or marketing, which was not the case.

Even Stanford computer science graduates “are struggling to find entry-level jobs” in “a dramatic reversal from three years ago”. Yet, it’s hard to find data that provides evidence linking AI adoption to reduced job postings, according to an article in The Stanford Review.

Meanwhile, Yann LeCun, former AI chief scientist at Meta and founder of AMI Labs, has criticised Anthropic’s Amodei for his remarks that AI will reduce tech jobs by 50%. Lecun noted that AI lab CEOs—including himself, OpenAI CEO Sam Altman, Yoshua Bengio, and “Godfather of AI” Geoffrey Hinton—aren’t necessarily the right authorities to forecast AI impact on jobs and the broader economy. Instead, Lecun noted that economists can be a better judge of how AI will affect the job market.

“Dario is wrong. He knows absolutely nothing about the effects of technological revolutions on the labour market. Don’t listen to him, Sam, Yoshua, Geoff, or me on this topic. Listen to economists who have spent their careers studying this.”

By 2030, Gartner says, CIOs will expect all IT work to involve AI, with 75% handled by humans augmented by AI and the remaining 25% performed autonomously. The research advisory maintains that AI is less about job loss and more about workforce transformation, and expects AI’s overall impact on global employment to remain neutral through 2026 and turn net-positive by 2027 as new roles outpace those displaced.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we unlocked today is based on Skills in Chrome.

What problem does it solve? Here’s a situation most of us don’t notice because we’ve accepted it as normal: you craft a great AI prompt that works perfectly—say, one that extracts key information from any product page. It works. Then the next day, new tab, new page—you type the whole thing again. Or you find yourself digging through old chat histories just to retrieve that prompt you wrote three weeks ago.

For a product manager evaluating vendors, this becomes a real productivity leak. Researching five competing tools means retyping the same prompt five times, across five tabs, every single time. In roles where the same analysis is repeated—procurement, legal review, competitive research—that friction quickly adds up to hours lost each week.

Skills in Chrome addresses this directly. Save a prompt once, and reuse it anywhere. Run it on any page, or across multiple tabs, with a single click.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: Open Chrome → Click the Gemini sidebar icon → Sign in with your Google account

Skills in Chrome can help you:

• Save and reuse your best prompts with one click, no retyping.
• Run one prompt across multiple open tabs simultaneously.
• Browse a readymade Skills library and customise to your needs.

𝙴𝚡𝚊𝚖𝚙𝚕𝚎:

A product manager is evaluating five vendor tools. She types the same prompt on every site: “Summarise pricing, core features, and integrations from this page.”

With Skills, she does this once—and never again.

1. Open Gemini sidebar → Type and run her prompt on any vendor page
2. At end of chat → click “Save as Skill” → name it “Vendor Snapshot”
3. Open all five vendor tabs
4. In Gemini sidebar, type / → select “Vendor Snapshot” → add remaining tabs via +
5. Hit run-structured summaries across all five tabs, instantly
6. What took 20 minutes now takes 90 seconds.

What makes Skills in Chrome stand out?

• Multi-tab execution: One prompt, five tabs, one click—no other browser or AI tool offers this today.
• Workflow memory built into your browser: Skills sync across all signed-in Chrome devices automatically.
• Privacy-first: Confirmation prompts before any sensitive actions like sending emails or calendar events.

𝙽𝚘𝚝𝚎: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS & BYTES

Humanoid beats human record at Beijing half-marathon

A humanoid robot developed by Honor has captured global attention after winning a half-marathon race for robots in Beijing, posting a time that surpasses the human world record for the distance. The event, held in the Beijing Economic-Technological Development Area (E-Town), highlights China’s accelerating progress in artificial intelligence and robotics.

The winning robot completed the 21-km course in 50 minutes and 26 seconds. This time is notably faster than the current human half-marathon world record held by Jacob Kiplimo, who ran the distance in approximately 57 minutes at a Lisbon road race in March 2026.

A separate Honor robot—operated via remote control—crossed the finish line even faster at 48 minutes and 19 seconds. However, under the competition’s weighted scoring system, which prioritised autonomy, the fully autonomous humanoid robot was awarded first place.

ChatGPT Images 2.0 ups the stakes

The internet is buzzing over OpenAI’s rollout of ChatGPT Images 2.0, which it frames as the start of “a new era of image generation”. The update builds on the December launch of ChatGPT Images positioned as a response to Google Gemini’s Nano Banana.

ChatGPT now “thinks” before it creates—planning outputs, pulling in references, and checking for errors before delivering results. The upgrade also brings practical gains: up to 2K resolution, batch generation of eight images, flexible aspect ratios from 3:1 to 1:3, and improved multilingual text rendering. It is not perfect, though.

I prompted it to “Please draw a picture of a Royal Enfield cruiser bike with essential accessories.” What I got was a decent picture but the fuel tank, as you see, belongs more to the Classic Model rather than the Meteor 350.

As ChatGPT Images 2.0 works to improve with time, the move intensifies competition in a crowded field that includes Midjourney, Stable Diffusion, Adobe Firefly and Google Gemini.

GPT 5.5 ‘Spud’ looms, GPT-6 still a meme

Reports and speculation suggest ChatGPT 5.5 (codenamed “Spud”) will be a modest upgrade within the GPT-5 family, aimed at bridging the gap between GPT-5.4 and GPT-6, with a possible release by April end. OpenAI, however, has made no official announcement.

As for GPT-6, the signal is even noisier. In October 2025, CEO Sam Altman leaned into the viral “6-7” meme, joking on X that GPT-6 might be rebranded “GPT-6-7,” after Dictionary.com crowned the absurd phrase its Word of the Year. The comment was tongue-in-cheek but underscored how Big Tech firms make gossip work for them.

You may also want to read:

• Apple may soon let you unlock your Tata car using iPhone
• Meta to track keystrokes, mouse movements for AI training
• ‘DeepSeek running on Huawei chips is a ‘horrible outcome’ for US’
• Amazon to invest up to $25 billion in Anthropic, deepening AI ties
• In the AI propaganda war, Iran is winning