Ciente > Blog > News

DeepSeek Shares Preview of their Highly Anticipated Model Designed for Huawei Chips

Category: News

Discover the latest in technology, from breaking news to trending updates. Ciente.io brings together all key tech stories.

DeepSeek
News

DeepSeek Shares Preview of their Highly Anticipated Model Designed for Huawei Chips

DeepSeek Shares Preview of their Highly Anticipated Model Designed for Huawei Chips

DeepSeek V4 is here to break the bank, not your budget. 🇨🇳 1.6T parameters, Huawei-powered, and 7x cheaper than Claude. Is the AI crown moving East?

If you thought the AI arms race was strictly a Silicon Valley affair, China just dropped a 1.6-trillion-parameter reality check. DeepSeek V4 is here, and it’s not just a model- it’s a geopolitical statement wrapped in code.

A year after they stunned the world by matching Western benchmarks at a fraction of the cost, DeepSeek is doubling down. The new V4 lineup, featuring Pro and Flash versions, is a technical marvel that shouldn’t technically exist given the export bans on NVIDIA’s high-end chips.

But here’s the nuance: DeepSeek didn’t just find a workaround; they pivoted to Huawei. By adapting V4 to run on Huawei’s Ascend hardware, they’ve effectively de-NVIDIA-ed their future, proving that compute-hungry AI can still thrive behind a trade wall.

The pricing is absolutely aggressive.

At roughly $3.48 per million output tokens, the V4-Pro isn’t just competing with Claude or Gemini; it’s attempting to bankrupt the concept of premium AI pricing. We are looking at a 7x price gap compared to Western flagships for performance that, on many benchmarks, trails only Google’s Gemini-Pro-3.1.

Is it perfect? Not quite.

DeepSeek is still fighting off heavy allegations from the White House regarding intellectual property theft. The timing of the launch- right after US accusations of industrial-scale IP theft- feels less like a coincidence and more like a flex.

For users, the real win is the 1-million-token context window. While others are still struggling with memory issues, DeepSeek is pushing a world where you can feed an entire codebase or a year’s worth of financial data into a single prompt without breaking the bank. It’s a shift from AI that “chats” to AI that “analyzes” at scale.

Whether you’re a fan of the company or a skeptic of the geopolitics, one thing is clear: the era of Western AI exceptionalism is officially under siege.

Google's
News

Pichai Unveils Google’s Roadmap at Google Cloud Next’26

Pichai Unveils Google’s Roadmap at Google Cloud Next’26

Google Cloud Next 2026 is a $185 billion bet on the agentic era. Is the human developer becoming obsolete, with 75% of code now AI-written?

We thought the AI hype cycle was starting to lose steam. But Google just dropped a $185 billion reality check. At Google Cloud Next 2026, Sundar Pichai didn’t just announce some new chips; he essentially declared the end of SaaS and the birth of the Agentic Enterprise.

Google is no longer interested in just giving you a chatbot to help you write emails. They want to give you a digital workforce. The headline-grabber is the Gemini Enterprise Agent Platform- a mission control designed to manage thousands of autonomous agents simultaneously.

But the real hold my coffee moment?

Pichai revealed that AI now generates 75% of all new code at Google. Let that sink in.

We’ve moved past the “AI as a co-pilot” phase and straight into “AI as the primary engine.” When the company that basically built the modern internet is allowing AI to write three-quarters of its stack, the human-in-the-loop narrative starts to feel more like a safety net than a requirement.

The nuance here, however, isn’t just in the solution- it’s in the silicon.

While everyone else is fighting over Nvidia H100s, Google is quietly building a vertical monopoly with its 8th-gen TPUs (TPU 8t and 8i). By owning the chips, the model, and the agent platform, Google is solving the compute-cost problem that is currently killing its competitors. They are selling the most efficient factory to run it.

The skeptic in me wonders: how many agents can a company actually govern before the reasoning drift becomes a liability?

Google is betting big on Agentic Defense to fix that, but we’re entering uncharted territory where businesses are autonomous- beyond automated.

Whether you’re ready for it or not, the agentic era is here.

Google isn’t just moving the goalposts; they’re rebuilding the entire stadium. And if your business plan still treats AI as a side-hustle rather than your core infrastructure, you’re extinct.

Microsoft
News

Microsoft Is Tired of You “Using” Office. They Want You To “Vibe” With It.

Microsoft Is Tired of You “Using” Office. They Want You To “Vibe” With It.

Microsoft’s Vibe Working is here. From static docs to autonomous agents, Office just got a major AI personality hire.

If that sounds like corporate gaslighting, you’re not entirely wrong, but the technical shift behind the buzzword is actually massive.

Microsoft’s new Agent Mode, rebranded as “Vibe Working,” is the company’s admission that the first wave of Copilot was, well, a bit of a letdown. It was a glorified search bar that occasionally hallucinated a bad paragraph. This new era? It’s about building a digital co-worker that actually stays in the room.

The vibe here is about persistence.

Traditional AI is one-and-done- you prompt, it answers, and the context is lost. Agent Mode flips the script.

The Excel agent doesn’t just suggest a formula- it detects a mess of date formats, proposes a cleaning pipeline, executes the fix, and then waits for your next move. It “speaks Excel” natively, i.e., treats your spreadsheet as a living environment.

But the real nuance lies in the behavioral layers.

You can now toggle vibes such as Formal Review or Creative Draft in Word. It moves us past the era of generic AI-speak. It’s an attempt to solve the uncanny valley of AI writing by allowing users to dictate the operational personality of the agent.

The funniest part of the whole announcement: Microsoft is hedging its bets.

Anthropic powers the Office Agent in Copilot chat, while the core Agent Mode is built on OpenAI’s reasoning models. It’s a wild crossover episode- proof that even Microsoft knows it can’t win the AI race with a single horse.

And honestly, “Vibe Working” does seem like a desperate attempt to make the drudgery of spreadsheets and slide decks feel cool again. But beneath the Gen-Z branding is a serious power play.

By turning Office into an agentic platform, Microsoft is essentially automating the “busywork” of middle management. If an agent can research, draft, and format a 20-slide deck while you’re getting coffee, the definition of “productivity” just changed forever.

You’re no longer a writer or an analyst; you’re a vibe manager.

IBM
News

IBM and ServiceNow Slip Turns the “Software-mageddon” Alarmingly Real

IBM and ServiceNow Slip Turns the “Software-mageddon” Alarmingly Real

Software-mageddon is here. As IBM and ServiceNow slip up, the market is asking if we still need SaaS in an AI agent world?

If the semiconductor boom is a gold rush, the traditional software industry just realized it’s the one selling the shovels. And the miners just found a way to manifest gold out of thin air.

On Thursday, the “software-mageddon” narrative went from a whisper to a scream.

Despite IBM and ServiceNow reporting numbers that technically beat analyst estimates, their stocks took a nosedive. IBM’s shares slid over 7% after growth in its Red Hat unit (usually the company’s crown jewel) stuttered. ServiceNow didn’t fare much better, flagging subscription hits and geopolitical friction.

But look past the spreadsheets, and the real story is much more existential.

Investors are no longer buying the “we have an AI story” pitch. They’re looking at the $1 trillion in market value that has evaporated from the software sector since January and asking a brutal question: In a world of autonomous AI agents, do we even need your software anymore?

The Anthropic effect is looming large here.

When Anthropic launched tools earlier this year that could automate complex data analytics and even modernize COBOL code, it directly threatened the sticky enterprise relationships companies like IBM have relied on for decades.

Why pay for a massive subscription to a platform that manages your workflows when an AI agent can do the work across your existing systems?

We are witnessing a violent rotation in the market.

The money is flowing out of “Software-as-a-Service” and into “Intelligence-as-a-Service.” If you make the chips (Nvidia, Texas Instruments), you’re winning. But if you’re a mid-level software giant whose business model relies on charging “per-seat” for a tool that humans use, you’re in the crosshairs.

The nuance is that software isn’t “dead”. But it has been demoted from the platform to the plumbing. The value is shifting from the interface to the inference. As one analyst put it, the challenge has moved from having an AI story to proving AI returns. If your software is just a mediator between a human and a task, your days are numbered.

The build era is enroute to changing what we’re willing to pay for- if not design in itself.

Meta's
News

Meta’s Employees are Now Its Very Own AI Training Data

Meta’s Employees are Now Its Very Own AI Training Data

Meta is recording every employee’s keystroke to train its AI. Is this frontier research or just high-tech surveillance? The digital sweatshop has arrived.

Think again if you thought corporate surveillance peaked with return-to-office mandates. Meta just took the Big Brother trope and turned it into a training manual.

According to a new Reuters report, Meta is launching the Model Capability Initiative (MCI), a program that installs software on U.S. employees’ computers to record every mouse movement, keystroke, and click.

The goal?

To feed that digital exhaust into their next generation of AI agents. Meta is asking its employees to help build their own automated replacements- by harvesting the muscle memory of their daily work.

Let’s get into the fascinating yet uncomfortable nuance here.

Anthropic is building tools to help you design. But Meta is cultivating tools to replicate the way you interact with a screen. Spokespeople are quick to promise that this data won’t be used for performance reviews- which, frankly, feels like being told the giant recording device in your living room is only for product research.

Even if we believe them, the irony is thick: while employees are being recorded to train Superintelligence, the company is simultaneously prepping for a 10% global workforce cut.

The technical justification is that current AI still sucks at the small stuff- the dropdown menus, the keyboard shortcuts, the rhythmic navigation of a complex UI. By capturing real-world trajectories, Meta hopes to bridge the gap from a chatbot that gives advice to an agent that actually does the job.

But here’s the real takeaway: we’ve officially moved past the era of training AI on public data. The open web has been picked clean.

Now, tech giants are turning inward, mining the very movements of their own staff to find the next competitive edge. It turns white-collar work into a sort of digital assembly line where your value isn’t just the code you ship, but the specific way your hand moves the mouse while you do it.

Meta calls it the “Agent Transformation Accelerator.” Most employees would probably call it a digital sweatshop.

Either way, the message is clear: if you work in tech, you aren’t just an employee anymore- you’re the data.

Is Musk Building an AI Empire? His $60 Billion Bet Makes It Seem So
News

Is Musk Building an AI Empire? His $60 Billion Bet Makes It Seem So

Is Musk Building an AI Empire? His $60 Billion Bet Makes It Seem So

$60 billion for a coding tool? SpaceX is eyeing a massive takeover of Cursor AI. Musk is building an AI empire, and your IDE is the new battleground. Read why.

Elon Musk doesn’t do small, and his latest power move makes that abundantly clear.

SpaceX currently has two options: either buy AI coding startup Cursor for a staggering $60 billion or drop $10 billion just for a seat at the partnership table.

Now is the time to wake up. Musk is building a “vertically integrated” AI ecosystem that owns the intelligent infrastructure. The topic of discussion is no longer Mars or satellites.

Cursor has become the darling of the dev world by making AI coding actually usable, but they’ve been relying on models from rivals like OpenAI and Anthropic.

By folding them into the SpaceX/xAI ecosystem, Musk is giving them the keys to “Colossus”- his massive Memphis-based supercomputer cluster. We’re talking about a million H100 equivalents. It’s like handing a world-class driver a jet-powered hypercar.

But let’s look at the why behind the $60 billion price tag. SpaceX is eyeing a $1.75 trillion IPO, and they need to prove they aren’t just a hardware play. By securing Cursor, they’re positioning themselves at the center of the developer productivity market.

If you own the IDE where the world’s best engineers work, you own the brain of the tech industry.

The real controversy is the talent grab.

Two of Cursor’s top engineers have already jumped ship to join SpaceX’s lunar projects. It’s more like a gradual assimilation.

This is a double-edged sword for an average developer. On one hand, the sheer computing power could make Cursor’s tools god-like. On the other hand, the tool you use to write your company’s secret sauce might soon be owned by a man who isn’t exactly known for playing well with others in the open-source community.

The coding wars have officially entered orbit- and the stakes just got exponentially higher.