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

The AI Engineer World’s Fair has ended, and over 7,000 folks are heading back to their homes, states,...

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.