Two good friends of mine in the past week have been laid off. Microsoft has announced it is cutting 60,000 jobs in an effort to reduce costs amid the rising use of AI. My Youtube feed is currently inundated with videos from Software developers who have just been laid off and don't know what to do next.

It's a scary prospect to any developer out there at the moment. Not only has your day-to-day job changed significantly in the past two years but now you have the real prospect of losing your income. You are becoming surplus to requirements.

Whether you think AI is the best thing since sliced bread or the worst thing to happen to Software since Windows 8 you likely still need a job. Software developers have suddenly found themselves in survival mode.

I want to help you stay in your job though so I've distilled what you need to do down to 5 tactics. These are tactics to use at work to make sure you not only survive but thrive in this AI world we find ourselves in.

Start Using AI Right Now

This may be obvious but if you're still sitting on the fence about AI it's time to make the leap. LLMs are not going anywhere in Software, if anything they're going to become the backbone of software production. So you need to make a decision.

If you've been an AI pessimist up until now it's time to change your tune. People notice. Your Boss notices. Whether you like it or not you will have built up a reputation.

It's time to change it.

This is about playing the game and changing the perceptions of you in your workplace. You don't have to be an AI diehard but you don't want to be known as 'The Complainer'. The person who whines and moans about 'the good old days', that LLMs 'hallucinate' and that they write 'crappy code'. That just isn't the case anymore.

Let's get one thing clear: in 2026 if you're using one of the latest models then they write decent code. More decent than most human developers if I'm being really honest. The whole hallucination argument is very weak these days. From my own experience at work and in personal projects I can categorically state that LLMs write good clean code. BUT and this is a big 'but', it's up to you as the developer to provide the environment, rules and conditions to enable them to do that.

Learn The Jargon

Another important part of using LLMs is learning the jargon. When speaking about LLMs at work you want to make sure you sound competent and you can keep up in work conversations.

You need to know the difference between a 'harness' and an 'agent'. A 'model' and a 'provider'. What 'tokens' are what a 'context window' is.

Then you need to know how they all fit together. What harnesses different developers are using and what are the pros and cons of each. Which models are good for which type of tasks. The devil is in the details in this one, it's quite easy to sniff out people who don't know what they're talking about.

Which leads into the next point...

Keep Up To Date

This is becoming somewhat easier recently as the rate at which models are coming out appears to be slowing. 18 months ago it seemed that every week a new revolutionary model was coming out and the FOMO was at an all time high.

For example Anthropic released its last flagship model - Opus 5.5 - on the 22nd September 2026. Due to its upcoming IPO it is delaying it's pre-training on the next model to tighten security and model weight protections. With the company going public it's suddenly under much more scrutiny around safety so it won't be able to release models as quickly.

This is a relief to many developers as they can finally settle on a workflow with a reliable model and not worry about suddenly being out of date.

But the world of LLMs still moves quickly compared to other industries so you must try to keep up. I'm not saying you always have to use the latest and greatest model being championed by the "AI gurus" on X, but you should at least be aware of them.

This is just the state of the industry at the moment whether you like it or not. So rather than fight it and waste your energy, accept it and focus your energy on becoming knowledgable in this domain. I personally use X, Syntax.fm and Medium to stay up to date.

Use LLMs as Collaborators

One thing that has really helped me is reframing how I think about LLMs. When they first came along I thought of them as competitors. I saw LLMs as competing with me for my job and that somehow I had to prove I was better than them.

This was an exhausting approach.

I eventually realised that LLMs are a tool that are incredibly good at some things and not so good at other things. Therefore, as the old adage goes, "If you can't beat 'em, join 'em".

This requires knowing which tasks in your job to hand off to LLMs and which tasks you keep to yourself and guard religiously. This will vary from person to person and from job to job. However, I have a general rule I like to stick to when making these decisions.

If it's a repetitive task that doesn't require much thinking but lots of text - Perfect for LLMs. If it's a one off task that does require thinking then I need to make those decisions.

As you can see this isn't a strict formula and there are many areas of grey but it does help me split tasks. Take for example "Creating a PR". I am creating many PRs throughout my day. It is repetitive. There is lots of text involved when writing the description, what to check, what requirements were covered etc. It's a perfect task to hand off to an LLM.

Whereas if I had the task to combine two separate applications into one unified interface, then this is a big one-off task that requires a mass of context. It requires information from stakeholders, users and product. Sure, I can hand off small tasks within this larger task to LLMs but I very much want to be making the big decisions.

Focus On The Architecture

As much as I think we need to embrace AI and LLMs we shouldn't hand over all responsibility. Not yet anyway.

If you think about an area or subject you know nothing about - lets say Architecture - LLMs will definitely be able to teach you something about it. However, they could very easily lead you down paths that are totally inappropriate to your current situation. And you wouldn't have a clue.

This is where expertise and experience comes in. If you were to take a suggestion from LLMs and ask an expert who has been in the field for 20 years on their opinion you are bound to find some differences. The more experience I have in Software the more I have learnt that making any decision requires trade offs. It requires considering both the short term and long term consequences.

Without my experience and knowledge these decisions become very hard and I would have to put more trust into the LLMs. This is dangerous ground as the LLMs have different goals and agendas to the developer. The developer is ultimately responsible for the piece of software. LLMs tend to go for quick fixes. They don't always have the most up to date context. They aren't always aware of the latest business strategies.

Gain as much knowledge as you can along the way. All of this will aid you when having to make your own decisions about software that you are responsible for. Then you can control and guide the LLMs to do the implementation and utilise them for what they are best suited for.

Conclusion

It is quite sad that I felt the need to write this article. It's giving advise to 'survive' in the current job market. This whole situation doesn't fill me with joy but I feel it's important.

Whilst some developers have adapted quickly to the use of LLMs in producing Software, many are struggling. The loss of flow, the loss of meaning and the loss of challenge are all well reported. Now however, real jobs are being lost and it's important that those who are struggling wake up and start doing what is needed.

So start using LLMs if you're not already. Start pushing the boundaries of what you think they can do, find the edge and stay close to it. But don't hand over everything to them, discern where your real value lies and where their value lies. Humans and LLMs can both enjoy being at this party.