DEV Community: Iain Thomson

Is looping ready to roll? Experts split on the future of coding

Iain Thomson — Fri, 03 Jul 2026 15:07:40 +0000

On the closing day of the AI Engineer World's Fair, industry leaders debated whether loops are ready for mainstream use and whether humans still have a place in coding.

On the loops-now side, Geoff Huntley, founder of Latent Patterns, was understandably bullish. He said that over two years ago he was at Canva and watched engineers prompting repeatedly and realized that this could be done much more efficiently with a software loop.

"It's somewhat inevitable," he said. "It is not a complete silver bullet. This time next year, at the conference we're going to see a whole bunch of talks saying our factories fail, our loops fail. These are things that we are yet to figure out."

He likened the situation to the first deployments of Kubernetes, where it took the industry years to get deployment right. But when they did, it was a revolution.

Dex Horthy, CEO of HumanLayer, was less enthusiastic. Earlier in the week, he gave a keynote in which he described how — as an experiment — he'd taken humans out of the coding loop and let machines do the job. He monitored the results, and they weren't good. He felt it was clear that AI wasn't up to the job yet.

"I think the basic take here is not whether loops are good or bad," he said. "I've seen lots of people try to apply AI to this problem of having to review bots, but it doesn't feel to me like it's working."

The increasing use of loops was inevitable, he said, but too many people are rushing into the space with too little thought. A more measured approach was needed, and some of the most enthusiastic early adopters would get their fingers burned.

The debate, done in the Oxford-style timed format, was closely fought, and the audience was asked to vote on either side. Unfortunately, the stage lights meant the moderator couldn't count votes, but — based on our view of the audience — Horthy's viewpoint won by a narrow margin.

The biggest World’s Fair yet powers down for the moment

Iain Thomson — Fri, 03 Jul 2026 14:45:31 +0000

The AI Engineer World’s Fair has ended, and over 7,000 folks are heading back to their homes, states, and foreign countries, and mulling the important lessons learned.

This has been the largest gathering of AI engineers since the conference began in a small way three years ago. Its new home in San Francisco’s Moscone West space oddly mirrors Google’s I/O developer conference, which started as a small gathering in Mountain View and three years later moved into the Moscone due to pressure of numbers.

It has been a novel trip for many. I witnessed three developers from overseas gazing into a parked Waymo with wonder, before the AI-powered car moved off to pick up its next passenger. One, rather sweetly, wished it a “bye bye,” and, based on the system’s popularity, the cars are fast becoming a tourist attraction.

But the conference was all about working and learning. While the keynote speeches garnered huge audiences, it was in the workshops that hardcore coding got done. The show hosted over 100 workshops, ranging from straight product talks to capture-the-flag sessions, roundtable discussions, and coding classes to make models more useful.

AI And The Jobs Dilemma

There was much discussion about the effect of AI on the jobs market, both in public and in private.

If you’re a skilled engineer then there’s no problem – recruiters are beating a path to your door. But for people in more traditional careers the outlook is mixed. Code reviewers should start upgrading their skills, since the increasing efficiency of AI can do their job at a speed and volume is something a human could never match. But there’s no need to switch careers – just move up the development stack and oversee the agent army to make sure it’s doing its job.

Despite talks of job losses, more than one CEO we spoke to is bullish – he’s hiring as many engineers as he can since AI is making them so productive they more than pay their way. He’s looking at AI as a way of expanding his number of employees as a win-win.

On the other hand, for ordinary workers, the outlook is mixed. One attendee said a client of theirs was hiring bank salespeople for eight months and mining their pitches to build an AI pitching agent, then letting them go. His comments brought to mind the long practice of hiring cheap H-1B workers and getting established engineers to train staff in the jobs they would shortly lose.

For hardcore coders the situation is looking less rosy for those who have coasted in their careers. During the show, the U.S. government relaxed its prohibition on Anthropic’s Fable 5 and Mythos 5. While the latter is more important for security engineers, Fable could be a serious threat to those who don’t know how to use it.

Getting Loopy

A key theme of the conference was loop coding in AI systems, and the debate has been constant and occasionally vociferous.

There has been much debate about loop coding – whether it saves money and tokens, if the loop can truly move from codebase to practical apps, if it will lead to more efficient working or if there are hidden problems under the surface, and if it is just an inevitable move for the industry.

That loops are coming doesn’t seem to be in doubt. Last month Microsoft's CEO Satya Nadella was quite clear that this is the next big thing, as was OpenClaw’s founder Peter Steinberger. But it remains to be seen if this is going to be an immediate thing, or a slow burner like Kubernetes or cloud computing.

It has been a fractious debate at times throughout the week. The key question seems to be how efficient looping will be and the cost involved. On the security side there is the fear that loops will be deluging security staff with AI slop – filing huge numbers of flaw reports that repeat common flaws and cause staff to miss more serious problems.

On the cost side the question is will loops save companies token costs, and the jury is still out on that. On one level users can own their own loop systems, which could be a cheaper option than going to the hyperscale vendors. However, if companies aren't prepared to commit to that, then loops could be a money pit.

The Importance Of Lobbycon

Meeting folks in the lobbies of the hall isn’t in the schedule but its value cannot be overstated.

The second floor of the center was abuzz with people catching up with old friends, making new ones, and sharing chat and code that built new relationships. There was even a talk on the topic – possibly the one that mentioned AI the fewest times during the whole show – but was important in building friendships.

Lunch was a key part of this. While the Moscone does sell food, large groups of people headed out to the many restaurants and watering holes in the vicinity to have in-depth conversations. One suspects that more than a few business plans were forged over the last week that will lead to startups shortly.

There’s also the fun side. With the World Cup in town there were more than a few football shirts in evidence – happy USA fans after their win, but the one chap in a Tartan Army shirt was visibly upset after Scotland’s finest failed to live up to the hype. Parties abounded in local venues, with people having fun and building contacts.

As delegates head out of the cool, grey city of love they’ll be carrying important knowledge, useful experiences, and the knowledge that San Francisco isn’t the hellhole it’s portrayed as in certain parts of the media.

Some Robots Just Can’t Handle The Expo

Iain Thomson — Thu, 02 Jul 2026 15:05:23 +0000

As you'd expect, there were robots aplenty at the AI Engineer World's Fair Expo, although with mixed results.

Many companies had little four-wheeled box robots, such as those used to deliver shopping in some cities. One would dispense small swag gifts in exchange for a QR code, while others just jerked around the floor, stopping and starting to avoid running people over.

Those that garnered the most attention were the humanoid bipeds, but this proved to be a mixed blessing. While they were attracting a lot of attention, it wasn't always the best kind. We observed numerous delegates having great fun trying to push them over.

This seems a tad rude and, judging from the picture, at least one was left broken in a tangle of limbs. Its compadre was all right, however, and was in the traditional kneeling pose while it recharged, or "praying to the great power socket," as one wag put it.

Google VP of Technology says he’s given up on coding

Iain Thomson — Thu, 02 Jul 2026 14:48:23 +0000

In his keynote on Wednesday, Benoit Schillings, vice president of Technology at Google DeepMind and formerly CTO at Google X, said he’s finally given up on coding his own software and has handed it off to agents.

Schillings said he was a bit new to DeepMind, having moved over from Google X only 18 months ago, and said the shift was “an interesting formative experience.” He explained that he’d been writing code for 45 years, inspired by the desire to write better Apple games, and said he was “pretty resistant to change” on that score. But that has changed.

“It’s about superhuman syntax generation. The last time I got Gemini to write a function for me, and I looked at it, and I was like, ‘do that better.’ Now it's over. I think that the minutiae of code writing can now be generated. Our time is gone.”

Humans still have a role, but it’s more in application and systems design. Humans are still better than machines at that — for now — but, with about 80% of code in GitHub now machine-generated, the era of the human code writer and checker is drawing to a close.

“I would predict that within one year we'll let Gemini or other models generate the code, and nobody will actually look at it, you know,” he told the crowd. “It's similar to compilers. Who still changes the assembly output of their compiler?”

He added that this shift would also require new programming languages to greatly increase the security of agentic code. They wouldn’t be perfect, he said, but we must improve from languages like Python.

Anthropic Fires Up Its New Engine

The speechwriters at Anthropic must have had a busy night rewriting presentations in light of the U.S. government’s decision on Tuesday to lift the ban on Fable 5 and Mythos 5, and Thariq Shihipar, a member of Claude’s technical staff at Anthropic, was in a boisterous mood for his keynote.

“It's one of those Anthropic models that you're just going to remember,” he enthused.

“The best way to describe Fable is like the map is opening up when you’re playing an RPG. You've been on the tutorial, and now you get to the point where the open world starts and there's so much that you can do and explore.”

This is the latest iteration of Claude Code, from its initial design to its current capabilities, including the ability to work proactively and reduce system prompts. He described the new build as “unhobbling Claude,” and said that developers will face a steep learning curve with the new build.

The full capabilities of the Fable build will be discussed in a separate article in this issue, since Shihipar and Claude’s head of code did a group session later in the day.

Amazon Ups The Perception Game

Antje Barth, who’s on the technical staff at Amazon’s AGI Lab, used her keynote to tout the benefits of the company’s latest "perception agent harness” that’s designed to increase efficiency in end-to-end workflows.

“We just launched the first two pieces of our Perception Agent Harness — open source,” she said. “There's two pieces. It has annotation, which you can use to tell it what you want and the verification part that gives the agent the capability to check its own work.”

She also demonstrated an annotation feature, a Chrome extension that allows users to select elements on the screen and mark changes. It logs human alterations, creates a complete summary, and applies the new rules, thus reducing the need for back-and-forth communication.

Fable Is Set Free - There’s A Brand New Claude In Town

Iain Thomson — Thu, 02 Jul 2026 14:18:20 +0000

As we initially covered yesterday, after some heavy lobbying by Anthropic, U.S. Commerce Secretary Howard Lutnick lifted the restrictions on Fable 5 and Mythos 5 after fears that they could be too powerful.

The following day, Thariq Shihipar, a member of Claude's technical staff at Anthropic, and Cat Wu, the company's head of product for Claude, held a roundtable hosted by Simon Willison, founder of the Datasette open-source project, to promote the new model and sing the praises of its efficacy.

The Anthropic duo claimed that Fable blew past competitor models in terms of speed and efficiency, with agentic coding and code fixing particular strengths. In the latter case, it was dealing with 50% more pull requests for its engineers.

The complexity of the code that Fable can create is better than anything Anthropic has managed before, the team asserted. It can operate for longer without needing human input, check its own prompts by writing test scripts for software it has created, and handle visual information from embedded graphs and images in source material.

"It's good at the dynamic permissions that you yourself give inside of the prompt, which I think is really important," Shihipar said.

"It also works well with our sandboxing instructions because sandboxing is one of those things where there's so many different edge cases, and it's hard to determine if we follow them. But if you have a sandbox and something needs to escape like a network request, auto load can then look at that request and allow it if it makes sense."

Complex data analysis capabilities have been beefed up, it was said, again without having to constantly question the human operator. Working within an agent harness like Claude Code, it can delegate work to subagents for greater autonomy.

"In general, we try to keep the cardinality pretty low and make sure that every tool we add has a distinct function from every other tool, so that Claude can very easily distinguish when to call each for file edit," Wu explained.

"The reason that we have file edit is because back in the day we used to show, or I guess we still do, we show people when Claude makes a file change, and there's this nice dedicated UI that says, 'Do you approve this edit to this file?'"

Wu recounted how she'd been asked to join a Microsoft Teams meeting for the first time and decided to see if she could get Fable to do it for her. She asked it to set up an account and add the necessary contacts to the group. It added guests, wrote a welcome message, and contacted IT admins about restricted accounts, all in 30 minutes.

"I love this idea that configuring Microsoft Teams takes an advanced AI 30 minutes, that's fantastic," commented Willison.

One feature Wu was particularly keen on was remote control. Fable can be set up for a long process and left to get on with it but still be controlled by a separate laptop or smartphone.

"A lot of people tell me that they just plug their laptop into a power charger, close the screen, open a bunch of remote control sessions, watch the screen, and then use their mobile phone from their couch to control Claude," she recounted.

It's not all work, though. Shihipar admitted he'd used it to develop a 2D fighting game and build characters of his friends. Wu has a personal project to support her mountain-climbing habit, with an application that finds routes to scale, matches them to local Airbnbs, and plots the distance to the slopes from the nearest parking space. Neither application is scheduled for release.

Visitors get some serious puppy love

Iain Thomson — Wed, 01 Jul 2026 13:47:36 +0000

There may be robots aplenty on the AIE expo floor, but many delegates have been drawn to a more mammalian exhibit. In the expo hall, the charity Puppie Love is letting visitors get some serious petting time with its young canine charges.

Visitors spend time with the dogs, but grabbing one is strictly forbidden. However, the puppies are keen to play, and the fluffball to the right of the picture should be named Hugging Face, given the amount of face licking going on.

Two of the puppies have been adopted, but a representative said the process is more difficult as many people are traveling to the show. Attendees can't just pop a pup into their swag bag, as the adoption paperwork understandably takes a few weeks.

The charity tours trade shows across the U.S. seeking adoptees for its young charges, with considerable success in finding happy families for those in need.

AI is going loopy, but in a good way

Iain Thomson — Wed, 01 Jul 2026 13:38:05 +0000

As you’d expect, the opening keynote of the AI Engineer World’s Fair was kicked off by one of its co-founders, @swyx (Shawn Wang), and he was in a poetic mood.

“In the beginning, there was the token, then there was the chat,” he said. “Then we're going to use tools, then we learn to set goals of skipping a few steps, and these days for all of the automations for all of the products. There's a lot of loops happening.”

At a basic level, AI agent loops work by having the system evaluate its own intermediate output — checking it against the task's success criteria or running it through an evaluator step — rather than simply returning the first response. If the evaluation indicates the task isn't complete, the system makes additional calls to the LLM, incorporating tool results or prior output, and repeats until the task is judged done, without needing a human to intervene at each step.

These loops can also compound over time: As employees correct and refine the system's outputs, those interactions can be captured to improve future performance — not just within a single task, but across the organization's use of the system.

Microsoft CEO Satya Nadella made this case in a LinkedIn post two weeks ago, framing it as something companies should own rather than cede to AI vendors: "This loop becomes the new IP of the firm. I think of it as a hill climbing machine. And unlike most assets, it compounds."

Unsurprisingly, Pablo Castro, a distinguished engineer at Microsoft, agrees with his boss. He claimed that as loops have been deployed within the company, they have produced major increases in the accuracy of AI outputs, and it has built a component called Agent Optimizer in its Azure AI Foundry platform for customers to use.

“We want to make it easy for you to integrate them into agents you're building,” he explained.

“We do this from our agent platform that starts in GitHub, where we all go and build as a contextualization system, so you can ground your agents. When it comes to hosting some mobility and management, we do all of these in Foundry. We offer thousands of models in our model catalog there, so you can pick whatever is the right model for the right task, and we will keep adding more every day.”

‘There has never been a better time to be an engineer’

During OpenAI’s keynote, Romain Huet, its head of developer experience, said that technologies like loops have dramatically increased productivity within the company. Previously, OpenAI was putting out new models every 15 months, but now it is taking six weeks.

“There has never been a better time to be an engineer, because engineering was never about writing code. Engineering has always been about solving problems for yourself and for other people as well,” he enthused.

“It's about taking the latest science and combining it with design, taste, judgment, and most of all, imagination to make something that people can actually use. We think it's a return to the roots of engineering and the technology we're building on is accelerating, getting faster and faster.”

Peter Steinberger, founder of OpenClaw, was brought on stage to also sing the praises of loops, saying they have completely changed the way he works. In January, Steinberger recounted, he was juggling 10 terminal windows at a time. Now he has a management system that clears away the easy ones and allows him to deal with harder problems.

This doesn’t entirely solve bottlenecks in a project, however, he explained. “Last year, I was primarily constrained by tokens — I fixed it by joining OpenAI,” which caused some audience mirth. “Now I'm primarily constrained by attention, and unlike tokens or compute, I can't simply add more of it. So the most important skill today is deciding where to spend it.”

Training loops are absolutely crucial, OpenAI’s Alexander Embiricos, the head of product for Codex, argued in his section of the keynote. Users need to train the system about not only the work that is required, but why it has to be done. The end results allow faster prototyping and give a major productive boost.

Speeding up the learning process

In his keynote, just to show that there were no hard feelings, Thom Wolf, co-founder of Hugging Face, brought on stage someone who he’d tried to hire, but who had turned him down. Olive Son is the research lead at MiniMax, a company Wolf described as “one of the top of what we call the AI dragons in China.”

Last month, MiniMax released M3, an open weight large language model that can work from text, image, and video inputs. Son said that the video angle was particularly important, claiming that M3 could actually process YouTube videos and learn from them.

“We know that a lot of labs run into problems doing that — the model would collapse after a couple of training steps with both text and visual understanding, but we managed to solve that problem,” she explained.

“We did a lot of work on VIP (Value-Implicit Pre-training), and we did a lot of work on the data that we're actually training. For example, we do what we call interleaved data — it's actually natural data — but we keep the images and videos in instead of masking it out, and we do some pretty good cleaning and masking on the data, and we do very good reward modeling, so that we train it from the first step and it scales up a lot.”

This kind of model was envisaged by Yan Junjie before the company even started, Olive said. It’s the most recent open weight multimodal model — meaning people can use it, but not access the training data, training code, or inference operators. But the support of the open source community was important, she said, and could make the model get better than it is now — and MiniMax can use that knowledge to improve future model builds.

Security track makes its debut

Finally, the first keynote session concluded with the announcement of an entirely new track for the Fair. Randall Degges, vice president of engineering and developer relations at AI security company Snyk, said the conference would now have a dedicated security track.

He acknowledged that there are security problems with the technology, but said that these are often overhyped — not least by governments. There were cheers from the audience when he cited the recent blocking of Anthropic's Fable 5 and Mythos 5, and the limited release of OpenAI GPT 5.6.

“As part of our ongoing engagement with the U.S. government, we previewed our plans and the models’ capabilities ahead of today’s launch,” OpenAI said, through gritted teeth, one suspects, last Friday. "At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly.”

Degges said the talk in the track should reassure many, as well as provide innovative new ways to use AI to lock out security gremlins.

Trust but verify when using AI for fixing security flaws

Iain Thomson — Wed, 01 Jul 2026 13:20:20 +0000

AI might seem like a magic bullet for fixing security issues, but it's not that simple, warned Eugene Yan, a member of technical staff at Anthropic, during the newly inaugurated security track at AI Engineer World's Fair. The effectiveness of AI in finding and fixing flaws is doubling every five months, he said, pointing to Mozilla releasing a 423-patch bundle in April. This was more patches than were released in all of 2025.

But while agents are good at finding and fixing flaws, the human element is still needed, say many security professionals. This is both to check that the AI has done the work properly and to make sure that seemingly low-risk bugs can't be strung together to make a serious exploit that AI might not spot.

To fix this, Yan proposed a six-stage program. "We found that most teams converge in approximately these six steps, and a big chunk of my thoughts will be about these," he told the crowd.

First, a threat-finding stage identifies a potential flaw and transfers it to phase two, a sandbox, to see if proof-of-concept code can exploit the issue. The third stage is a discovery phase in which the sample is checked against past issues that may have been fixed.

Stage four is an independent verification, which is designed to further filter out false positive results, and then the results are triaged to avoid flooding out human checkers. Then a patch is developed, and the code is kicked back to the discovery engine.

The end result, he argued, will be much more secure code that still maintains human oversight — while making the lives of security staff a lot easier. Of course, as AI systems improve further, that may not be the case forever if the current rate of engine improvements continues.

It’s Time To Put Humans Back In The Software

Iain Thomson — Wed, 01 Jul 2026 13:19:01 +0000

Software engineers have become overreliant on models to build applications, and it’s time to put humans more firmly in the loop, the AI Engineer World’s Fair was told on Tuesday.

The term software engineering was first used in a 1968 NATO conference, which was called to discuss the problems of building large-scale, reliable applications at speed. It was suggested that applying industrial manufacturing to code creation would help speed this process up, but with larger and larger amounts of code now being machine generated, the cracks are starting to show in reliability.

“We are all racing to put AI coding into production, and there's been lots said about loop engineering, and we should probably write more loops,” argued Dex Horthy, the impressively mustachioed co-founder of HumanLayer.

“I'm here to convince you today that this is, in fact, not a joke, and that no amount of harness engineering or loops maxing can solve what is fundamentally a model training issue.”

The problem, he said, is that developers have become overreliant on badly trained models. While some agents are great at some parts of the development cycle, they are lousy at others. As a result, pull requests are up, code breaks are multiplying, and customers — and the companies supplying them — are losing out.

His proposal is to make humans much more involved in the early stages of software design, particularly with planning and architecture decisions. Human steering at the start can solve a lot of problems before they become serious issues, he opined, but 30 minutes more in preplanning and alignment can save hours in review.

Loops don’t offer a magic bullet, Horthy stated — it’s not possible to throw engineers into training agentic models to do their jobs for them. Loops can help, but they aren’t the universal panacea that many in the industry are saying.

Part of the problem is the reward system for training, Horthy said. If a serious flaw is committed in vibed code, it could only be discovered months or even years after the error was made. There’s no way to “propagate that reward signal back across the gap,” he told the crowd.

Human code reviews are too often being skimped on, or even eliminated, in some software houses. AI can help human checkers considerably, but it can’t be relied on without human oversight, he proposed. And if you’re getting drowned out in AI-generated pull requests, then that also indicates the models are at fault. He added that if machine-made code needs a 20% rewrite, then that’s actually pretty good for much of the agentic output.

Small fixes can be assigned to an agent to fix, but for key system architecture decisions, humans need to review and plan where a product is going, how its parts fit together, and how to get there without too many pull requests. Mistakes will always be made, but it’s important to minimize them if you want to keep users from getting overly pissed off. The answer is a fusion of man and machine, but with some realignment, the issue is fixable, he suggested.

“It's easy to hear all this and be a little bummed out. I really like the world where we can just not have to ever read code ever again,” he said.

“But we're engineers, and these are just constraints, and models are good at certain things, and they're not good at other things. And so, go figure out how to solve problems given a set of constraints. Use loops, they're great. But go solve the hard problems, seek leverage.”

AI Drone Swarm Shines Out Over San Francisco

Iain Thomson — Tue, 30 Jun 2026 15:15:34 +0000

San Franciscans and attendees at the AI Engineer World's Fair saw a shining swarm of drones flying over the iconic Ferry Building waterfront last night.

While America loves its fireworks, particularly on the upcoming July 4 "Huzzah, we kicked out the Brits and will spell color any way we please" celebration, these are increasingly seen as so last century. Instead, AI-controlled drone swarms are the coming thing, and LLM code creation and curation company Qudo put on a show for those who looked up on Monday evening.

The event was held in the gourmet's paradise of the Ferry Building and drew engineers from around the world to talk about AI and coming trends in the industry. Then a swarm of battery-powered drones took to the sky for a flyby.

While this might seem like a trivial use of the technology, it's increasingly welcome in the U.S., as the nation suffers an intense drought and the risk of wildfires needs to be controlled.

Hackathon Winners Scoop $35,000 In Cash And Credits

Iain Thomson — Tue, 30 Jun 2026 14:52:38 +0000

After a marathon 48 hours of coding, the winners of the AI Engineer World's Fair Hackathon have been announced.

In third place — winning $1,000 in cash, a further $1,000 in MiniMax credits, $5,000 in Atlas AI credits, and 200 million Voyage AI tokens — was Team Rote. The crew claimed that they'd been able to make an AI engine that was the fastest on the planet by collecting agent usage and making it available to multiple people across the network.

Coming in second was Team PodMan, which built a tool that orchestrates tasks using a single agent on DigitalOcean servers, managing data in real time to prevent merge errors and improve efficiency by storing and reusing executed commands using a specialized MongoDB database. As such, the 500 million Voyage AI tokens will be very helpful with further use of MongoDB, and the other prizes of $1,500 in cash, $2,000 in MiniMax credits, and $5,000 in Atlas AI credits will be welcome too.

But the judges were most impressed with Team SplatForge, which clinched the top spot using Google Gemini to identify and move objects in virtual environments in 3D. Key to its success was that the software learned to find and classify new objects. The ability for systems to learn and develop was a key point behind the hackathon and something the judges were specifically looking for. The hacking session, held at the offices of the Shack15 social club in the San Francisco Ferry Building, was clearly grueling. One developer was asleep on a couch by the time the final presentations were made. We didn't wake him to check which team he was on, but since he was up and about by the time the awards were made, we hope he had a good wake-up call.

Scientists On AI: It’s Still Experimental

Iain Thomson — Tue, 30 Jun 2026 14:39:00 +0000

As the AI Engineer World's Fair Expo kicked off, a group of scientists gathered for an informal get-together to discuss how this new technology is affecting their profession.

"I work in chemistry, so hallucinations are still a big problem," organizer Scott Reed, a professor at the University of Colorado Denver, told us. "But the bigger picture I've opened up is that science@AIE provides a better forward face for AI, that there's benefits, and that people don't only see the downside."

Generally, the news is positive. When it comes to coding research apps, AI can be a boon, so long as the code is checked carefully, since scientific notations are sometimes misconstrued or mistaken.

With physical testing, there's still a way to go, however. As one academic from the University of Southern California who develops turbines put it, "Put the bolt in the wrong place, and the engine will sling a blade into the next state."