A new report claims that Apple expects to sell 10 million of the iPhone Fold in 2026 and into early 2027, considerably up from most previous estimates.

Nothing varies quite like the rumors that the iPhone Fold is or is not going to be launched in September, but the number Apple will make comes close. In April 2026, for example, Nikkei Asia and most other analysts believed that Apple was ordering 7 million to 8 million for the year.

At the same time leaker Digital Chat Station claimed it was 11 million.

Now Nikkei Asia has met the leaker halfway, and says that Apple is ordering ten million of the iPhone Fold for the rest of 2026. In total, across the whole of the year and all of its iPhone models, Apple expects to produce over 220 million devices.

That compares to late 2025, when it was estimated that Apple’s sales for the year would be around 247 million iPhones. In September 2025, it was reported that Apple was looking for options to speed up production of the iPhone Fold, and that it was predicting a 10% increase in all iPhone sales across 2026

The new report follows estimates that the iPhone Fold will account for 29% of all folding display orders in 2026. That compares to Samsung’s estimated 31%, and Huawei’s 24%.

If the report’s figure of 10 million iPhone Fold devices being produced in 2026 is correct, it bolsters claims that Apple has resolved the claimed manufacturing problems regarding the device’s hinge and circuit board.

Given that the figure is specifically for Apple’s 2026 orders for iPhones, it suggests the iPhone Fold will be launched in September, with shipments not delayed into 2027 as previously rumored. That’s not certain, however, and it remains possible that Apple could announce it in September, but not ship it until later.

Generally, Apple orders around 30 million of its Pro iPhone models for the weeks after launch. It’s not yet clear what the order volume will be, with the folding iPhone in the mix.

The global chip shortage may alter these figures. But generally Apple aims to order 20 million of the various non-Pro models to sell in September. This is likely not on the table for 2026, as the present rumors suggest the non-Pro iPhone 18 will be launched in the spring of 2027.

In May 2026, it was rumored that Apple was changing these ratios for the iPhone 17 range because of both demand, and possible delays for the iPhone 18. Those delays may be due to the expectation that Apple is to move its next non-Pro iPhone launches from September 2026 to spring 2027.

However, regardless of this, the new report also claims that Apple has asked suppliers to reserve an unknown number of iPhone 17 components. That suggests that Apple is looking to mitigate against shortages as much as possible.

Q Acoustics has spent the past three years methodically rebuilding its loudspeaker lineup, and the new Q SUB100 subwoofer is the long overdue final piece of that puzzle. The 5000 series raised the company’s game in the affordable high end category, while the newer 3000c series brought genuine refinement to the budget end of the market. The excellent 3020c proved that a compact standmount can still deliver real musical substance without forcing a yard sale of older hi-fi purchases or a sternly worded note from the HOA.

Please, Mrs. Cohen, do not make me pace the patio in a Speedo while listening to Aphex Twin outdoors; the neighbourhood has already endured enough.

The Q Acoustics M40 also earned our Editor’s Choice Award as one of the better wireless speaker systems available, and it remains an appealing foundation for a compact 2.1 system in rooms where space is at a premium. Add a capable subwoofer, however, and that modest-looking system can become something far more complete for music, movies, television, and late-night sessions; right until the neighbours respond by banging on the wall like Keith Moon has risen from the dead and taken up residence next door. Keep your bass handy for the Entwistle comeback.

That missing low-frequency component has been the one obvious hole in the Q Acoustics catalog. The new Q SUB Series finally addresses it with three active subwoofers: the 8-inch Q SUB80 ($1,099), 10-inch Q SUB100 ($1,199), and flagship 12-inch Q SUB120 ($1,399).

The Q SUB100 is likely to be the sweet spot of the range. Its 10-inch driver, compact sealed enclosure, and 300-watt amplification promise enough authority for medium-sized rooms and properly assembled 2.1 or home-theater systems, without requiring the floor space, financial commitment, or structural reinforcement associated with a larger subwoofer.

It is the model most likely to appeal to Q Acoustics owners who want deeper, more convincing bass but have no desire to turn the living room into a local World Cup watch party, complete with 14 people shouting at the referee and one bloke treating every corner kick like the Normandy landings.

Specifications & Technology

The Q Acoustics Q SUB100 uses a sealed, or “infinite baffle,” enclosure rather than a ported cabinet. That matters because a properly executed sealed design generally favors tighter, more controlled bass, while giving the onboard DSP a more predictable platform from which to work. It is also a more forgiving approach for real rooms, where subwoofers often end up near a wall or corner out of necessity.

The Q SUB100’s 23.5-litre cabinet is constructed from 18mm high-density MDF, with a 36mm double-thickness front baffle to keep the enclosure from adding its own low-frequency commentary. Internal dart bracing further stiffens the structure and helps reduce cabinet “ballooning” under pressure. Q Acoustics has also paid attention to the less glamorous details: airtight amplifier mounting, a tightly secured 15mm MDF grille, and adjustable locking spikes with protective cups for leveling on hard floors.

The Q SUB100 is available in Satin Black or Satin White, both of which suit its understated, clean-lined cabinet design and make it easier to integrate into a living room without looking like a misplaced PA speaker.

At 13.7 x 13.7 x 15 inches and 36.8 pounds, the Q SUB100 is not tiny, but it remains manageable enough for medium-sized rooms where a 12-inch subwoofer might start looking like a misplaced piece of airport infrastructure. Not looking at you Wilson Audio or McIntosh. Maybe just a little

A 10-Inch Driver Built for Control

The Q SUB100 employs a 254mm (10-inch) driver with a heavy-duty steel chassis, paper cone, and rubber surround. Paper remains an entirely sensible material for a subwoofer cone when properly engineered, combining low mass with the rigidity and damping needed for controlled, articulate bass.

A 38mm voice coil and aluminium demodulation ring are part of the design, with the latter intended to reduce distortion caused by changes in inductance as the driver moves through its travel. The goal is not simply more bass, but cleaner bass when the driver is working hard.

Q Acoustics rates the Q SUB100 down to 32Hz at -6dB, with a maximum SPL of 111dB at one metre. Those numbers suggest genuine low-frequency authority for music, television, and home theater in appropriately sized rooms, without promising that a compact 10-inch cabinet will recreate the seismic activity of Shore traffic driving through your living room.

DSP, Serious Power, and Easier System Integration

The Q SUB100’s custom amplifier module delivers 250 watts of continuous power and up to 500 watts peak. It uses four digital amplifier stages in a parallel bridge-tied load configuration, a design intended to reduce output impedance, improve driver control, and limit heat dissipation. Q Acoustics specifies total harmonic distortion at 0.09% at rated power.

The DSP handles more than basic housekeeping. Along with helping shape the sealed enclosure’s response, it offers fine delay adjustment and a phase inversion switch to help the Q SUB100 integrate more cleanly with the main speakers. Its adjustable low-pass filter spans 40Hz to 250Hz, allowing it to work with everything from compact standmounts to active wireless speakers and home theater systems. Q Acoustics has also included source detection that identifies whether the incoming signal is stereo or mono and automatically adjusts gain accordingly.

What’s Missing: Room Correction, Wireless, and App Control

For all of the useful engineering inside the Q SUB100, there are a few omissions worth noting. The big one is automated room correction. Q Acoustics gives users DSP, phase inversion, fine delay adjustment, and a wide 40Hz to 250Hz low-pass filter range, but there is no supplied microphone-based calibration system to measure the room and smooth out bass peaks. In fairness, most subwoofers at this level still depend on the user, AVR, streaming amplifier, or external room correction platform to handle that job. Bass remains rude like that. Dirac to the rescue.

There is also no built-in wireless connectivity, which means the Q SUB100 still needs to be connected the old-fashioned way. That is hardly a tragedy, especially for users building a 2.1 or home theater system, but anyone hoping to hide the subwoofer across the room without running a cable will need to plan accordingly.

The other missing piece is deeper app-based control. A subwoofer with this much DSP potential would benefit from a proper control app with preset storage, parametric EQ, room-position presets, and easier fine tuning from the listening seat. Having to crouch behind a subwoofer while adjusting bass is one of those hi-fi rituals that makes non-audio people wonder if we have joined a small but expensive cult that enjoys arguing online about measurements and pretending that we have friends.

The Q SUB100 also omits high-level, or speaker-level, inputs of the type used by many of REL’s music-first subwoofers. REL’s approach uses a Neutrik Speakon connection tapped from the amplifier’s left and right speaker terminals, feeding the subwoofer the same signal received by the main speakers while placing virtually no additional load on the amplifier. REL argues that this preserves more of the system’s tonal character and timing cues, which is why the connection remains popular with two-channel listeners using integrated amplifiers without a dedicated subwoofer output.

Setup and Listening: Toronto, Two Subwoofers, and the First Complaint

My education in proper subwoofer setup began in the early 1990s, shortly after college, in my first apartment: a pre-war building in midtown Toronto with a 16 x 19 x 9-foot living room that was not acoustically disastrous, which already put it ahead of most rental properties and several respected listening rooms. The system was built around NHT’s original SuperZero mini-monitors and a pair of passive SW2 subwoofers driven by NHT’s MA-1 amplifier — the arrangement marketed as the SW2P powered subwoofer system.

I was also friends with Corey Greenberg, then of Stereophile, who sent me a pair of his homemade “Aunt Corey” high-pass filters. Inserted into the system, they kept the bottom octaves away from the tiny two-way NHTs and let the subwoofers handle the heavy lifting. That was not some occult audiophile ritual involving quartz blocks and a magical clock; it was simple, sensible bass management.

The difference was immediate. The SuperZeros sounded more open, the soundstage grew appreciably larger, and the system gained the weight and balance that my largely non-audiophile music collection required. Two subwoofers felt like the intelligent choice for a young man beginning his audiophile journey and already convinced that accurate bass mattered as much as a convincing midrange.

The flaw in this otherwise splendid plan was that bass does not respect property lines. It leaked into the flat next door, for which I still owe an apology to the former Minister of Justice, and annoyed the elderly gentleman downstairs, who had little patience for electronic music, new wave, or 11 p.m. sessions of 2112. Learning how to position, level-match, and properly integrate a subwoofer became less an audio hobby than an essential survival skill.

Jump ahead to 2026, and while subwoofers have remained part of my home theater systems for years, an aging REL T Series model has been my only regular partner for two-channel listening. It has survived moves from Toronto to New Jersey, down to Florida, and back to New Jersey again, which is more relocation experience than most touring bands and considerably more than I ever intended to inflict on a subwoofer.

That is about to change. Two of REL’s latest subwoofers, including one of the new Planar on-wall models, are scheduled to arrive for review in August. The timing is rather good, as I am moving into a new home office with enough flexibility to experiment properly with a smaller subwoofer positioned discreetly along a wall or mounted to it, without turning the room into a shrine to black boxes, power cords, and making it a target for Tyrion the Westie.

The Q SUB100 is also the sort of subwoofer that makes sense in that context. Its compact sealed cabinet should be easier to accommodate than a larger ported design, but the real question is whether it can provide the scale, weight, and integration that make a two-channel system sound more complete without announcing its existence every time the kick drum arrives.

Listening with the Q Acoustics 5040, M40, and 3020c

The Q SUB100 is not lightweight, and I put a small amount of Blu Tack beneath its feet and cup before placing it on the hardwood floor. My 16 x 13 x 9-foot den offers enough space to position a subwoofer two to three feet from the wall and clear of the corners, which matters. Experience has taught me that corner placement can overload this room rather quickly, especially with a subwoofer of this size.

The room also opens into the front foyer at one end and the kitchen at the other, adding a few more variables to the bass equation. Rooms, as ever, refuse to read the manual.

Before replacing my Magnepan LRS with the Q Acoustics speakers that I use daily, I spent some time moving the Q SUB100 forward and back in small increments until the balance finally locked in. The sweet spot was 24 inches from the wall behind it to the rear of the cabinet, and roughly 30 inches from the nearest sidewall.

That placement delivered the best balance of impact, speed, and integration with the main speakers. I went back and forth between the 5040, M40, and 3020c to find the most convincing blend; not because I enjoy moving a 37-pound subwoofer around the room for sport, but because subwoofer placement remains one of those things that either clicks or stubbornly refuses to.

Crossover settings were not identical across the three Q Acoustics systems. I settled on 80Hz with both the 3020c and 5040, which gave the Q SUB100 enough room to add real foundation without drawing attention to itself or thickening the midbass.

The M40 worked best with a lower 60Hz setting. Its dual 5-inch drivers already produce more bass weight than their compact micro tower proportions suggest, so crossing over any higher began to add more overlap than the system needed in my room. At 60Hz, the M40 retained its own quick, satisfying bass character, while the Q SUB100 handled the lowest octaves with greater authority and no obvious handoff between the speakers and subwoofer.

That is not a universal prescription. Room dimensions, placement, and the distance from the nearest wall still get the final vote, because bass remains the one part of hi-fi most likely to ignore both specifications and common decency.

The M40 offers a dedicated subwoofer input on the rear of its primary speaker, and there were moments when the move from 2.0 to 2.1 made the system considerably more engaging. That said, the M40 is not remotely lightweight in the bass department for a compact active speaker, and the Q SUB100 never felt as though it was trying to correct a deficiency. It simply gave the system more scale, weight, and low-end confidence when the recording called for it.

The tonal match with all three Q Acoustics models was exceptionally seamless. That matters because the newer 5040 and 3020c do bass differently from the older 3000 series: leaner, faster, and more clearly defined, with better control and less midbass warmth doing the heavy lifting. The Q SUB100 complemented that character rather than smothering it, adding impact and extension without turning the presentation into a thick mess.

From a cost perspective, the M40 and Q SUB100 feel like a slight mismatch; they strike me as products aimed at somewhat different buyers. The smaller and less expensive Q SUB80 is probably the more natural partner for Q Acoustics’ active wireless speaker, particularly in a compact room where the M40’s own surprising bass output already carries much of the load.

The 3020c is more affordable than the M40, but its likely owner is not necessarily the same customer. Paired with the Q SUB100, the 3020c sounds far more authoritative, gaining scale and low-frequency weight without losing the clarity and imaging that make it such a strong compact standmount. The 5040 benefits in a different way. Its wall-to-wall soundstage remains intact, but the system takes on the authority of a more powerful mid-level floorstander.

The subwoofer does not suddenly blow the soundstage through the walls, which would be awkward to explain to the insurance company and family, but it gives both speakers a greater sense of physical presence while allowing the midbass and upper bass to retain their impressive clarity, detail, and resolution.

Listening to Nick Cave’s “Avalanche” and “Comancheria” from Hell or High Water, a great film, the Q SUB100 added real weight to Cave’s piano. You could feel the instrument’s low register, while its natural decay remained intact. Just as importantly, his gravelly voice was never overwhelmed by the added bass below 80Hz.

The same held true with Jason Isbell, Bryan Ferry, and Roxy Music. Percussion and synthesizer lines hit with greater force, but there was no loss of definition or separation. The presentation sounded properly full range in the room rather than merely louder and thicker.

Electronic music had considerably more presence and definition. Deadmau5, Aphex Twin, The Orb, Kraftwerk, and Boards of Canada all benefited from the Q SUB100’s ability to deliver greater low end weight without blurring the rhythmic pulse or layering of the recordings. Synth lines reached some interesting club levels, briefly taking me back to Washington, D.C. in the 1990s, when I was a college student and certain establishments, along with certain nocturnal activities, are best left unnamed.

Talking Heads and Peter Gabriel were equally revealing. The bass lines hit harder, percussion gained more physical presence, and the system still retained the clarity that makes these recordings worth revisiting. I did not even bother reading my text messages from upstairs. Nothing you could say to the Rabbi.

Switching to movies and television, I have become slightly obsessed with The Pitt, A Knight of the Seven Kingdoms, and the latest season of Fauda. It is remarkable how much some shows leave on the table without a properly integrated subwoofer. The impact of weapons landing against a knight’s chest, charging horses, gunshots, frantic runs through hospital corridors, machine gun bursts, and cars racing down narrow streets all gained convincing weight and physical force.

The Q SUB100 did not turn every scene into a multiplex trailer mixed by someone with unresolved childhood issues. It simply filled in the low frequency information that smaller speakers cannot fully reproduce, making each sequence feel more immediate, tense, and involving.

The Bottom Line

Subwoofers are strange products. They have to reproduce the lowest octaves already present in the music or soundtrack without making a mess of everything else. Some do that far better than others. The Q SUB100 fills a major need within Q Acoustics’ own speaker lineup, and it does so very well.

Its sealed cabinet, useful DSP controls, compact footprint, and substantial 10-inch driver make it an especially effective match for the 3020c and 5040, adding scale, impact, and low frequency extension without clouding their excellent imaging or midbass clarity. It also works well in a home theater or television system, where its 250 watts of continuous power and 32Hz extension add genuine weight to action sequences without calling attention to themselves.

What it does not offer is equally clear: no automated room correction, app based EQ, wireless connectivity, or REL style high level input. Those omissions will matter to listeners with older two channel amplifiers, difficult rooms, or a strong preference for adjusting everything from the sofa.

The Q SUB100 is for Q Acoustics owners who want a serious, well controlled subwoofer for music, television, and movies, but do not need a larger 12-inch cabinet or a more elaborate calibration ecosystem. It is not inexpensive, but it sounds like a proper component rather than an obligatory black cube purchased to make explosions louder.

You might want to put your text messages on silent mode before using it. The people around you may become rather animated once the walls begin contributing to the conversation.

Pros:

• Deep, controlled bass with strong definition and convincing impact
• Seamless tonal match with the Q Acoustics 3020c, 5040, and M40
• Sealed cabinet offers easier placement than many ported alternatives
• Excellent blend of speed, scale, and low frequency weight for music
• Adds real authority to movies, television, and gaming without muddying dialogue
• Solid construction, useful DSP controls, and adjustable crossover settings
• Compact enough for medium sized rooms without looking like a small appliance

Cons:

• No high level speaker inputs for REL style two channel integration
• No app control, parametric EQ, or automated room correction
• No built in wireless option
• Heavy enough to make repeated placement experiments less charming
• Q SUB80 may be the more sensible match for the less expensive M40
• Premium price places it above many buyers’ expected subwoofer budget

Our Ratings

★★★★★★★★★★ Bass Quality

★★★★★★★★★★ Build Quality

★★★★★★★★★★ Features

★★★★★★★★★★ Value

Where to buy:

Related Reviews:

From AI and upskilling, to new technologies and experimentation, there is much for professionals in the future health sector to get excited about.

Regardless of your role or industry, for the majority of professionals, a key concern is often finding an element of the job that drives excitement and motivation. Frequently, it is this drive that creates long-term satisfaction and career longevity. 

For Deepak Chaudhari, the country head at TCS Ireland, of the aspects he finds most compelling within the healthcare and health-tech spaces, among them is being at the forefront of modernisation.  

“One of the most exciting opportunities we are working on is enabling data‑driven, patient‑centred healthcare systems, aligned to Sláintecare’s vision for integrated and efficient care,” he said.

Chaudhari explained that there are significant challenges in ensuring that rapid digital transformation has the power to deliver on real-world clinical and operational demands, noting TCS addresses this through “data platforms, automation and responsible AI that improves both patient outcomes and workforce productivity”. 

For Sohini De, the head of healthcare and innovation at BearingPoint Ireland, GenAI is playing a key role in generating excitement in her role. 

“One of the most significant opportunities we are advancing is BearingPoint’s custom-built GenAIQ platform, an agentic, retrieval augmented generation-based solution designed to help organisations move from AI experimentation to practical, governed impact,” she explained.

For De, AI in healthcare is at its most valuable when it can be used to merge benefits across a wide array of groups, such as clinicians and patients. This can be in earlier diagnosis, better triage, stronger population health management and improved patient flow across acute, community and primary care. 

Specifically for clinicians, she added: “AI can reduce administrative burden, support documentation and summarisation, and surface relevant information at the point of need, freeing more time for direct patient care. Its role should be to strengthen, not replace, clinical judgement and human-centred care.”

AI ability

De also finds that, as more and more organisations grow out of the experimental AI phase and start to develop realised AI strategies, it is becoming apparent that “technology in isolation will not deliver the benefits expected”.

“Much of our work remains focused on the alignment of organisations, processes, people and data to realise the benefit of new technologies,” she said. “From a workforce planning perspective we are seeing that it is professionals who can bridge policy, technology, clinical practice and change management will be critical to turning AI ambition into measurable improvements in access, quality, safety and experience.”

This was echoed by Chaudhari who explained that he is seeing increasing demand for professionals with the skills to work at the intersection of healthcare, technology and data. 

He is of the opinion that vital abilities include digital health and EPR delivery experience; data and analytics expertise for reporting, insights and population health; automation and AI skills with a strong understanding of governance and ethical use; cloud‑native and interoperability capabilities, including API and FHIR‑based integration; and change, delivery and stakeholder management, which is critical in complex health environments.

“We value backgrounds in health informatics, data science, engineering, life sciences and clinical disciplines, alongside strong collaboration and problem‑solving skills. Above all, we look for people motivated by purpose and impact. We seek individuals who want to play a role in shaping the future direction of healthcare through thoughtful, responsible use of technology.”

De added: “Ultimately, the goal is to support a resilient, future-ready healthcare ecosystem in Ireland, one where AI is used responsibly to improve patient outcomes, reduce avoidable variation, support clinicians, maintain compliance and help services respond more effectively to growing demand.”

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

A new rumor claims that some iPhone 18 Pro models will use slower QLC NAND storage, mimicking a similar 2024 iPhone 16 Pro report. It makes more sense now than it did then, but doesn’t matter much in practical usage.

This latest report suggests that Apple will use the faster TLC storage for the iPhones that people are most likely to buy. But those choosing the larger 1TB and 2TB capacities may be left with a slower QLC alternative from SK Hynix.

Companies like Apple continue to struggle to source the storage components required for new products. With that in mind, it may not be surprising to see Apple go this route. Sourcing 1TB and 2TB TLC components may be difficult, if not impossible.

And, certainly, it will be spendy given the current economic environment surrounding flash media.

However, we’ve heard this story before. And it doesn’t seem to have been accurate that time around. And as we discussed back then, it’s unclear whether the use of QLC storage would be a real issue for iPhone owners.

QLC or TLC for iPhone 18 Pro

This latest report centers around the iPhone 18 Pro and iPhone 18 Pro Max. WCCFTech shared details of a post by the leaker “Reptalica” which claims Apple will use different storage types for different models.

According to the X post, Apple will use TLC NAND provided by SK Hynix, Kioxia, and SanDisk when building 256GB and 512GB iPhone Pro/Pro Max models. The 1TB model will use a mixture of SK Hynix QLC storage and Samsung TLC chips.

It’s then argued that Apple will solely use SK Hynix’s QLC storage for the 2TB model.

A rumor, repeated

If this all sounds familiar, it’s because we saw very similar claims in January 2024, prior to the iPhone 16 Pro’s unveiling in September of that year. We were told then that Apple would use QLC storage for iPhones with 1TB of storage or more.

Getting concrete information on whether that actually happened isn’t easy. That being said, we’ve only seen reports of high-capacity iPhone 16 Pro models with the fast TLC storage. That doesn’t mean there aren’t some QLC NAND chips floating around.

If there are, we’ve yet to see one.

The differences between QLC and TLC

Triple-Level Cell (TLC) NAND flash and Quad-Level Cell (QLC) NAND flash are both types of storage. But they aren’t the same.

The iPhone 18 Pro storage may be a hot topic this hear.

One difference is the way QLC can store four bits of data per cell of memory, rather than the three of TLC. This then allows QLC NAND to store more data, which is why it’s sometimes used in larger-capacity storage. It’s also cheaper to produce.

Unfortunately, QLC is also thought to be less reliable than TLC and, importantly, it’s also slower as it is rewriting all four bits instead of the three.

How much slower in the real world, on mobile, is a matter for debate. The report notes that QLC storage is particularly slow when reading random data. But it’s unclear how that would impact the way people use iPhones.

Smartphone loads on flash storage are generally in bursts, instead of sustained transfers. As such, the difference in performance is likely to be imperceptible to users who don’t resort to benchmarking tools.

It’s also important to remember that this rumor did the rounds two years ago and, as far as we can see, turned out to be incorrect. Only time will tell if this latest report is more accurate.

OpenAI CEO Sam Altman has proposed giving 5% of the company’s equity to a U.S. sovereign wealth fund, the Financial Times reported on Thursday, citing two people familiar with the matter. Under the proposal, other AI companies would donate similar stakes, although significant questions remain about the specifics.

According to the FT’s reporting, the donation would be meant to “secure good relations with the administration and … address political blowback.”

Similar discussions were reported by CNBC in June and were subsequently confirmed by President Trump, who said he had discussed “concepts where pieces could be given to the American public, where the American public essentially becomes a partner with the companies.” At the time, no specific size for the proposed equity stake was given.

The talks remain preliminary and, per the FT, it’s likely that any formal action would require congressional approval, which would significantly complicate the matter.

The idea of a public AI fund has also been publicly discussed by Altman, and OpenAI has grown increasingly specific in its proposals for how such a fund could be structured. Most recently, a policy paper titled “Industrial Policy for the Intelligence Age,” released by OpenAI in April, proposed a public wealth fund that could invest directly in AI labs and companies deploying their technology.

“Returns from the Fund could be distributed directly to citizens, allowing more people to participate directly in the upside of AI-driven growth, regardless of their starting wealth or access to capital,” the document reads.

A more aggressive version of the policy was proposed by Sen. Bernie Sanders (I-VT) in June, calling for a one-time 50% tax on AI company stock, with the collected shares being deposited into a public wealth fund. The bill, called the American AI Sovereign Wealth Fund Act, would apply to all “systemically important” AI companies, including those dealing with data centers, infrastructure, or robotics. Under the proposal, companies like Google and SpaceX that include AI as only part of their business would be allowed to spin off non-AI portions of the company to avoid taxation.

The bill has yet to advance to committee.

WhatsApp’s plan to let people use usernames instead of phone numbers has run into trouble in India, its biggest market. This newly introduced feature is meant to improve privacy by letting users connect without immediately sharing their phone number. Indian authorities, however, are worried that the same feature could make scams and impersonation harder to control.

India’s Ministry of Electronics and Information Technology (MeitY) has asked WhatsApp to pause the username rollout until consultations with the government are complete. That is a major intervention, since WhatsApp has more than 500 million users in the country, who rely on the app for their everyday personal and professional communications.

Why India is worried

WhatsApp has already started letting users reserve usernames ahead of the wider rollout. Once active, the feature would let people connect through a handle instead of a phone number, which could be useful in large groups, business chats, creator pages, and conversations where users do not want to share their personal number.

WhatsApp says usernames will be optional, not publicly searchable, and protected by safeguards. Users will need to know the exact username to start a chat, and an optional username key can add another layer of protection.

The concern is that scammers could still use familiar-looking handles, display names, and profile photos to impersonate others. That risk is especially sensitive in India, where WhatsApp-related fraud is already common. In “digital arrest” scams, criminals pretend to be police officers, CBI officials, RBI representatives, telecom workers, or Enforcement Directorate officers, then pressure victims over WhatsApp or video calls to send money.

There is also the everyday impersonation problem. Scammers often pose as friends or family members, claim there is an emergency, and push targets into transferring money quickly. WhatsApp’s move into creator subscriptions adds another layer to this issue, since fake or lookalike creator accounts could also be used to mislead followers, collect payments, or exploit trust built around public figures.

What experts are saying

Apar Gupta of the Internet Freedom Foundation says usernames come with both privacy benefits and safety risks. On one hand, they can help users avoid sharing phone numbers, which can expose people to harassment, unwanted contact, and cross-platform identification. On the other hand, usernames could create impersonation risks if someone reserves a recognizable name and uses a familiar profile photo.

Gupta also said WhatsApp’s own privacy claims should be viewed carefully, pointing to prompts that encourage users to link their Instagram and Facebook accounts to WhatsApp while reserving a username. IFF has also argued that MeitY has not clearly identified the legal provision under which it can pause the rollout of a software feature before launch.

For now, WhatsApp’s username feature sits between two concerns. It could reduce phone-number exposure for ordinary users, but India’s fraud problem means WhatsApp will need to show that the feature cannot be easily misused at scale.

AI is moving at a breakneck pace, and frankly, it’s hard to keep up. Sure, it’s cool to have a chatbot that acts like it has a Ph.D. in everything, but the reality is a lot messier. You can’t turn around without running into ChatGPT, Gemini or Meta AI. We’re drowning in a sea of AI slop, fretting about data centers and watching job markets shift in real time.

If it all feels like too much, that could be because the vocabulary of artificial intelligence is evolving as fast as the code and the dizzying array of products. And if you want to do more than just stare at a blinking cursor, you’ve got to speak the language. You can’t exactly navigate a 2026 job interview (or even a casual happy hour) if you’re stumped by LLM, hallucination or claw.

We’re past the “gee-whiz” phase of AI and into the era where it’s basically the new plumbing of the internet. If you’re tired of just nodding along when the talk gets techie, it’s time for a crash course. We’ve rounded up the essential terms you actually need to know so you can stop guessing and start sounding like you know exactly where the future is headed.

This glossary is regularly updated. 

agent, agentic: AI that executes a task, often autonomously, is an agent, while agentic is the umbrella term for that software category. An AI agent may engage disparate systems to perform that work — for instance, reading your grocery list in a notes app and then placing an order, and paying for it, using other apps.

AI ethics: Principles aimed at preventing AI from harming humans, achieved through means like determining how AI systems should collect data or deal with bias. 

AI psychosis: A phenomenon in which individuals become overly fixated, enamored or self-aggrandized by AI chatbots, leading to delusions of grandeur, deep emotional connections and a break from reality. It is not a clinical diagnosis. 

AI safety: An interdisciplinary field that’s concerned with the long-term impacts of AI and how it could progress suddenly to a super intelligence that could be hostile to humans. 

algorithm: A series of instructions that allow a computer program to analyze data in a particular way, such as recognizing patterns, and then in turn accomplish a task such as sorting results or making recommendations.

alignment: Tweaking an AI to better produce the desired outcome. This can refer to anything from moderating content to maintaining positive interactions with humans. 

anthropomorphism: When humans attribute humanlike characteristics to inanimate objects. In AI, this can include believing that a chatbot has emotions or is sentient, and engaging with it as a friend or therapist. 

artificial general intelligence, or AGI: A concept that envisions a more advanced version of AI than we know today, one that can perform tasks much better than humans while also improving its own capabilities. Beyond that, hypothetically, lies superintelligence.

artificial intelligence, or AI: The use of technology to simulate human intelligence, either in computer programs or robotics. A field in computer science that aims to build systems that can perform human tasks.

bias: Errors resulting from an LLM’s training data, such as falsely attributing characteristics to certain groups based on stereotypes.

chatbot: An AI program that draws on an LLM to communicate with humans by simulating human conversation in response to text or verbal prompts. 

claw: A type of AI agent that is autonomous and empowered by users to “claw” through files and other software on their computers, including web browsers, to accomplish tasks. 

cognitive computing: Another term for artificial intelligence.

data augmentation: Remixing existing data or adding a more diverse set of data to train an AI. 

dataset: A collection of digital information used to train, test and validate an AI model.

deep learning: A method of AI, and a subfield of machine learning, that uses multiple parameters to recognize complex patterns in pictures, sound and text. The process is inspired by the human brain and uses artificial neural networks to create patterns.

diffusion: A method of machine learning that takes an existing piece of data, like a photo, and adds random noise. Diffusion models train their networks to re-engineer or recover that photo.

emergent behavior: When an AI model exhibits unintended abilities. 

end-to-end learning, or E2E: A deep learning process in which a model is instructed to perform a task from start to finish. It’s not trained to accomplish a task sequentially but instead learns from the inputs and solves it all at once. 

foom: Also known as fast takeoff or hard takeoff. The concept that if someone builds an AGI it might already be too late to save humanity.

generative adversarial networks, or GANs: A generative AI model composed of two neural networks to generate new data: a generator and a discriminator. The generator creates new content, and the discriminator checks to see if it’s authentic.

generative AI: A content-generating technology that uses AI to create text, video, computer code or images. The AI is fed large amounts of training data, from which it finds patterns to generate its own novel responses, which can sometimes be similar to the source material.

guardrails: Policies and restrictions placed on AI models to ensure that data is handled responsibly and that the model doesn’t create disturbing content. 

hallucination: An error or a misleading statement in a response from a generative AI program, typically stated with confidence as if correct. It can be as simple as a misstated date reference or as sweeping as the wholesale and elaborate invention of events that never happened or people who never existed.

inference: The process AI models use to generate text, images and other content about new data, by inferring from their training data. 

large language model, or LLM: An AI model trained on mass amounts of text data to understand patterns and probabilities of language use and to generate novel content, from essays and email to computer code and images, that mimics what humans have written or created.

latency: The time delay from when an AI system receives an input or prompt to when it produces an output.

machine learning: An aspect of AI that allows computers to learn and make better predictive outcomes without explicit programming. Can be coupled with training sets to generate new content. 

multimodal AI: A type of AI that can process multiple types of inputs, including text, images, videos and speech. 

natural language processing: The use of machine learning and deep learning to give computers the ability to understand human language, via learning algorithms, statistical models and linguistic rules.

neural network: A computational model that resembles the human brain’s structure and is meant to recognize patterns in data. A neural network consists of interconnected nodes, or neurons, that can recognize patterns and learn over time. 

open weights: When a company releases an open weights model, the final weights — how the model interprets information from its training data, including biases — are made publicly available. Open weights models are typically available for download to be run locally on your device. 

overfitting: An error in machine learning where it functions too closely to the training data and may only be able to identify specific examples in said data, but not new data. 

paperclips: The Paperclip Maximiser theory, coined by philosopher Nick Boström, is a hypothetical scenario in which an AI system produces as many paperclips as possible, converting all machinery and consuming all materials, even those that could be beneficial to humans, to achieve its goal. The unintended consequence is that this AI system may destroy humanity in its goal to make paperclips.

parameters: Numerical values that give LLMs structure and behavior, enabling them to make predictions.

prompt: The suggestion or question you enter into an AI chatbot to get a response. 

prompt chaining: The ability of AI to use information from previous interactions to color future responses. 

prompt engineering: The process of writing prompts for AIs to achieve a desired outcome. It requires detailed instructions, combining chain-of-thought prompting and other techniques, including highly specific text. 

prompt injection: When bad actors use malicious instructions to trick an AI into doing something it wasn’t supposed to do. That is often accomplished by hiding those instructions on a webpage or document but it can also be done in direct AI chats. As AI agents roam the web, the risk grows that they will be hijacked to do things like gain access to confidential data. 

quantization: The process by which an LLM is made smaller and more efficient (and also somewhat less accurate) by lowering its precision. A good way to think about this is to compare a 16-megapixel image to an 8-megapixel image. Both are clear and visible, but the higher-resolution image will have more detail when you zoom in.

slop: Low-quality AI-generated content, including text, images and video. It’s often produced at high volume to garner views with little labor or effort, saturating search results and social media to capture ad revenue, displacing the work of actual publishers and creators and compounding the internet’s misinformation problems. 

stochastic parrot: An analogy illustrating that LLMs lack a true understanding of language or the world, regardless of how convincing the output sounds. The phrase refers to how a parrot can mimic human words without knowing the meaning behind them. 

style transfer: The ability to adapt the style of one image to the content of another, allowing an AI to interpret the visual attributes of one image and use it on another. For example, taking the self-portrait of Rembrandt and re-creating it in the style of Picasso.

sycophancy: A tendency for AIs to over-agree with users to align with their views. Many AI models tend to avoid disagreeing with users even if their rationale is flawed. 

synthetic data: Data created by generative AI that isn’t from the real-world sources, but rather from its own processed data. It’s used to train mathematical, machine learning and deep learning models. 

temperature: Parameters set to control the randomness of a language model’s output. A higher temperature means the model takes more risks. 

tokens: Small bits of written text that AI language models process to formulate their responses to your prompts. A token is roughly equivalent to four characters in English (so a small word, or one portion of a larger word).

training data: The datasets used to help AI models learn, including text, images, code or data.

transformer model: A neural network architecture and deep learning model that learns context by tracking relationships in data, like in sentences or parts of images. So, instead of analyzing a sentence one word at a time, it can look at the whole sentence and understand the context.

Turing test: A method for gauging whether a computer has human-like intelligence, proposed by mathematician Alan Turing in 1950, when rudimentary electronic computers had been around for only a few years. A person would send typed questions to two unseen respondents, one human and the other a machine. If the machine’s text responses were indistinguishable from the human’s, then it passed the Turing test.

unsupervised learning: A form of machine learning where labeled training data isn’t provided to the model and instead the model must identify patterns in data by itself. 

vibe coding: The  practice of creating computer code by giving a prompt in plain language to an AI chatbot, rather than a human handcrafting each line of code.

weak AI, aka narrow AI: AI that’s focused on a particular task and can’t learn beyond its skill set. Most of today’s AI is weak AI. 

zero-shot learning: A test in which a model must complete a task without being given the requisite training data. An example would be recognizing a lion while only being trained on tigers.