Four of the five technologies on this list have moved from conference slides into the real world. Humanoid robots now move totes and car parts in working facilities, more than 20 people use brain implants to control computers, and an optogenetic gene therapy for blindness is under FDA review. The fifth pick, the DNA-based molecular computer, never left the lab, so it has been swapped for a technology with better evidence behind it: error-corrected quantum computing.
The original list grew out of the Global Future 2045 congress at Lincoln Center in New York in June 2013, where roboticists, neuroscientists and geneticists pitched ideas about merging humans and machines. Earlier versions of this article dated that event to 2016. What follows checks each idea against what has actually shipped, been tested in people, or stalled.
Updated, September 14, 2026: Every section was rewritten with 2025 and 2026 evidence. Molecular (DNA) computers were replaced with quantum error correction because DNA computing produced no practical systems, while quantum hardware hit verifiable milestones. The congress date was corrected, a dead link was removed, and FAQs were added.
The 5 technologies at a glance in 2026
| Technology | Status, September 2026 | Names to know |
|---|---|---|
| Humanoid robots | Commercial pilots in warehouses and car plants; home robot sold to early buyers | Agility, Figure, Boston Dynamics, 1X, Tesla |
| Bionic limbs | Muscle-controlled arms on sale; nerve-integrated legs in small studies | DEKA / Mobius Bionics, MIT, Brigham and Women’s |
| Brain-computer interfaces | Human trials; no permanent implant approved for sale | Neuralink, Synchron, Precision, Paradromics |
| Optogenetics | First vision therapy filed with FDA | Nanoscope, GenSight |
| Quantum error correction | Below-threshold demos; fault-tolerant machines targeted for 2029 | Google, IBM |
1. Humanoid robots: from lifelike androids to warehouse workers
In 2013 the headline act was Hiroshi Ishiguro of Osaka University, who brought an android copy of himself on stage. He still builds humanlike machines: his Future of Life pavilion at Expo 2025 Osaka filled three zones with about 30 avatars and robots, including androids designed to look and act like people. Those remain exhibits and research platforms, not products.
The commercial money went elsewhere, to humanoids that look like machines and do repetitive work. That is a very different bet from the parkour videos of Boston Dynamics’ old hydraulic Atlas, and the numbers are now public:
- Agility Robotics Digit: passed 100,000 totes moved at GXO’s Flowery Branch facility in November 2025, lifting totes on and off mobile robots and conveyors in a live fulfillment operation.
- Figure AI: its Figure 02 robots spent 11 months at BMW, logging more than 1,250 hours of runtime and loading over 90,000 sheet-metal parts for more than 30,000 BMW X3s. Figure also disclosed that the forearm was the robot’s top hardware failure point, which is the kind of detail that separates a deployment report from a demo reel.
- Boston Dynamics Atlas: the electric Atlas entered production in January 2026, with the entire 2026 run committed to parent company Hyundai and Google DeepMind. It has 56 degrees of freedom, lifts up to 50 kg and can swap its own batteries. Other customers are slated for 2027.
- Tesla Optimus: Elon Musk promised 10,000 robots in 2025. In April 2026, Tesla said Optimus production at Fremont would start in late July or August on converted Model S and X lines, with no volume target.
Humanoids have also reached homes, with a large caveat. 1X sells its NEO for $20,000 or $499 a month, and for chores the robot has not learned, owners can schedule a remote 1X teleoperator to guide it. The company opened a 58,000 sq ft factory in Hayward, California in April 2026 with capacity for 10,000 robots a year.
Who should care: logistics and manufacturing buyers, because tote handling and parts loading are proven tasks now. Households should treat NEO as an early-adopter product: a person you have never met may be steering a camera-equipped robot around your home, and many of its promised skills depend on future software updates.
2. Bionic prosthetic limbs controlled by muscles and nerves
The “Luke arm” built by Dean Kamen’s DEKA received FDA marketing authorization in May 2014 as the first prosthetic arm able to make multiple powered movements at once. Earlier versions of this article said it is run by a foot joystick. That is only one option: the LUKE arm, now sold by Mobius Bionics, can be driven by surface EMG electrodes that read muscle signals, pressure switches or motion sensors worn on the feet, and it offers up to 10 powered joints from hand to shoulder.
Surface electrodes only listen to muscles through the skin. The bigger shift since 2019 is surgery that wires the prosthesis into the body’s own nervous system. At MIT and Brigham and Women’s Hospital, researchers reconnect muscle pairs during amputation, a procedure called the agonist-antagonist myoneural interface (AMI). In a 2024 Nature Medicine study of seven AMI patients, people walked faster than those with conventional below-knee amputations, at roughly the pace of people without amputations. The sensory feedback they regained was still under 20% of normal, and only about 60 people worldwide had received the surgery at the time.
In 2025 the same group added a titanium implant anchored in the thigh bone that carries nerve-muscle signals straight to a robotic knee. The two people using this osseointegrated mechanoneural prosthesis climbed stairs and stepped over obstacles better than comparison groups, but the researchers estimate FDA approval is about five years away.
Cost and comfort still decide whether people use these devices at all. A Dutch economic study put the mean yearly cost of a multi-grip myoelectric hand at €54,112, versus €23,501 for a standard myoelectric hand, and cited research showing that 4% to 50% of people with upper limb loss reject their prosthesis. For someone choosing a device today, a well-fitted myoelectric or body-powered arm and a good prosthetist matter more than headline research. Related wearable robotics is moving faster in industry, as powered exoskeletons for factory workers show.
3. Brain-computer interfaces: Neuralink, Synchron and the first human trials
A brain-computer interface (BCI) records electrical activity from neurons, usually in the motor cortex, and software translates those patterns into a cursor movement, a robotic arm command or text. In 2019 this was mostly university research. In 2026 it is a race between venture-funded companies running FDA-approved trials.
Neuralink moved from the 2021 demo of a monkey playing Pong to people. By January 2026 it reported 21 participants enrolled in trials worldwide, using implants to operate computers and an assistive robotic arm. One participant exceeded 10 bits per second on a cursor test in his first week, in the range of an able-bodied person using a mouse, and some participants have typed at up to 40 words per minute. The company reports zero serious device-related adverse events. The record is not spotless: in its first patient, many of the implant’s electrode threads retracted from the brain weeks after surgery in 2024, cutting performance until software changes recovered it.
Competitors take different approaches, and the differences matter for risk:
- Synchron threads its Stentrode through a blood vessel instead of opening the skull. Its six-patient COMMAND study reported no serious adverse effects on the brain or blood vessels, and the company is working with the FDA toward a pivotal trial that could lead to the first premarket approval of a permanent BCI. It raised a $200 million Series D in November 2025 to fund that work.
- Precision Neuroscience won FDA 510(k) clearance for its Layer 7 cortical interface in April 2025, a thin electrode film that sits on the brain’s surface. The clearance covers implants of up to 30 days, mainly for brain mapping during surgery, not permanent use.
- Paradromics completed the first implant in its Connect-One study on June 17, 2026, in a woman with motor neuron disease, aiming to restore speech.
Academic groups still post some of the strongest results. A UC Davis BrainGate participant with ALS has used his speech BCI at home for nearly two years, reaching 99% word accuracy and about 56 words per minute in a June 2026 Nature Medicine report. For now, implanted BCIs are for people with severe paralysis or ALS who enroll in trials. A consumer brain chip is not on any credible near-term timeline.
4. Optogenetics: from lab tool to the first vision therapy filed with the FDA
Optogenetics makes specific cells respond to light. Scientists deliver a gene for a light-sensitive protein, originally borrowed from algae, into target neurons, then switch those neurons on or off with flashes of light. Ed Boyden, Karl Deisseroth and colleagues showed millisecond-precise optical control of neurons in a 2005 Nature Neuroscience paper, and the method became a standard tool in brain research, mostly in animals.
The first human use is in the eye, because the retina is easy to reach with light. In 2021, a man blinded by retinitis pigmentosa partially recovered visual function after GenSight’s optogenetic gene therapy, perceiving and locating objects while wearing light-stimulating goggles.
Nanoscope Therapeutics has gone further. Its MCO-010 (brand name MOGENRY) is a one-time injection into the eye that makes surviving bipolar cells light-sensitive, and it does not require patients to have a specific gene mutation. In the RESTORE trial, 7 of 18 treated patients gained at least 0.3 LogMAR at 52 weeks (about three lines on an eye chart), rising to 10 of 18 at 76 weeks. Eye inflammation (44% versus 22% on sham) and raised eye pressure (39% versus 11%) were the common side effects, managed with eye drops. On September 9, 2026, Nanoscope said the FDA had accepted its Biologics License Application. No decision date has been published.
Hearing may be next. A University of Göttingen team building an optical cochlear implant planned a first-in-human safety study for 2026 and approval in the early 2030s. Whether that study has started is unconfirmed. People with advanced retinitis pigmentosa should follow the FDA review closely, with realistic expectations: the trial gains are meaningful, not a return to normal sight. For a playful, no-surgery take on crossed senses, the DIY Synesthesia Mask lets wearers smell colors.
5. Quantum error correction (replacing molecular computers)
The 2013 pitch was that computers made of DNA would outperform silicon. That did not happen. DNA’s strongest computing use turned out to be storage, and even there Microsoft Research, which has studied it with the University of Washington since 2015, says DNA data storage is not practical yet because synthesis and sequencing lag behind. DNA logic circuits remain lab experiments, so this slot now goes to a post-silicon computing idea with measurable progress.
Quantum computers use qubits that are extremely error-prone. Error correction spreads one reliable “logical” qubit across many physical qubits, but that only helps if adding qubits removes more errors than it introduces. Google’s 105-qubit Willow chip crossed that line in December 2024: as its test grid grew from 3×3 to 5×5 to 7×7 qubits, the error rate was cut in half at each step, a result published in Nature. In October 2025, Google reported a Quantum Echoes algorithm running 13,000 times faster on Willow than the best classical method, tested on molecules with 15 and 28 atoms.
IBM’s roadmap targets Starling, a fault-tolerant machine with 200 logical qubits and 100 million gates, in Poughkeepsie, New York, by 2029, using codes IBM says need about 10 times fewer qubits than the surface codes Google uses. In July 2026, IBM and partners including RIKEN and the University of Chicago published three papers claiming verifiable quantum advantage, though on scientific benchmark problems rather than business workloads.
The part that affects everyone is encryption. Google’s Craig Gidney estimated in 2025 that 2048-bit RSA could be broken in under a week by a machine with fewer than a million noisy qubits, down from 20 million qubits in his 2019 estimate. No machine is close to that size, but the estimate keeps falling. NIST finalized its first three post-quantum encryption standards in August 2024 and urged organizations to start migrating immediately. If you manage systems that store long-lived sensitive data, that migration is the practical takeaway from this section.
How to judge the next “amazing technology” claim
The 2013 predictions that aged best were the ones tied to measurable engineering. The ones that aged worst came with a date for replacing the human brain. Long-range forecasts rarely age evenly, as a scorecard of Ray Kurzweil’s predictions shows. A few checks help sort signal from hype:
- Regulatory status: “FDA accepted an application” and “FDA approved” are years apart. Check which one a headline means.
- Deployment numbers: hours, units and patients count for more than demo videos. Figure’s 1,250 hours and Neuralink’s 21 participants are useful because they are specific.
- Disclosed failures: companies that publish problems, such as retracted electrode threads or failing robot forearms, are usually further along than those that publish none.
- Who pays: a technology with a paying customer, like warehouse humanoids or the drones now used for delivery and inspection, tends to improve faster than one waiting for a sponsor.
Emerging technologies FAQs
What are the 5 amazing technologies you should know about in 2026?
The five covered here are humanoid robots, bionic prosthetic limbs, brain-computer interfaces, optogenetics and quantum error correction. Quantum error correction replaced DNA-based molecular computers, which never produced practical systems.
How many people have a Neuralink brain implant?
Neuralink reported 21 participants enrolled in its trials worldwide in January 2026. They use the implant to control computers and, in some cases, an assistive robotic arm. The device is still investigational and not approved for sale.
Can you buy a humanoid robot for your home in 2026?
Yes, with caveats. 1X sells its NEO humanoid for 20,000 dollars or 499 dollars a month, with US deliveries planned for 2026. For chores it has not learned, a remote 1X operator can guide the robot, which raises privacy questions.
Is optogenetics approved for use in humans?
Not yet. On September 9, 2026, Nanoscope Therapeutics said the FDA accepted its application for MCO-010, an optogenetic gene therapy for retinitis pigmentosa. No FDA decision date has been published.
Are mind-controlled prosthetic arms available today?
Muscle-controlled arms are. The DEKA LUKE arm received FDA marketing authorization in 2014 and reads muscle signals through skin electrodes. Prostheses wired directly into nerves and bone are still in small studies, with MIT researchers estimating about five years to FDA approval.
Can quantum computers break encryption yet?
No. A 2025 estimate by Google researcher Craig Gidney says breaking 2048-bit RSA would need close to a million noisy qubits, far more than any current machine. NIST published post-quantum encryption standards in August 2024 so organizations can switch before that happens.
Share Your Thoughts