Last20
banner
last20.bsky.social
Last20
@last20.bsky.social
1 followers 44 following 39 posts
Stuck on the Last 20% of your AI/no-code build? Post your task. Get it finished by a dev who knows your stack. No posting fee. Pay when the job is done.
Posts Media Videos Starter Packs
5/ Find a developer who specialises in finishing AI projects
- They understand the common patterns
- They won't judge your AI-generated code
- They can work with your existing structure

Stop trying to become a full developer.

Start focusing on shipping your idea.
4/ Set a realistic budget
- Most "last 20%" fixes cost £50-100
- Less than you'll spend on wasting your time trying to fix things.
2/ Document exactly what's broken
- Screenshot the error
- Copy the exact error message
- Note what you expected to happen

3/ List everything you've already tried
- This saves the developer time
- Helps them avoid duplicate work
How to get unstuck when AI tools fail you (learned the hard way): 🧵

1/ Stop throwing more AI credits at the problem
- If 3 different AI tools can't solve it, it's not an AI-solvable problem
- You're just burning money at this point
8/ Finally posted on Last20. A developer fixed it.

9/ Fixed in a few hours.

That's why we built Last20. For moments exactly like this.
4/ Bought Cursor Pro (£23) thinking a different AI would help

5/ Burned through £31 in Claude Code calls asking for debugging help

6/ Paid for Stack Overflow Premium (£15) to ask better questions

7/ Bought a Udemy course on deployment (£29)
Story time: How I spent £200 trying to fix a £20 problem 🧵

1/ Built 90% of my MVP with Lovable. Felt like a genius.

2/ Hit a deployment issue. "How hard can this be?"

3/ Spent £47 on additional Lovable credits trying different approaches
Interesting shift I'm seeing: More and more founders are building 80% of their apps with AI tools, then getting completely stuck.

They're not looking for full dev teams or 6-month projects.

They need someone to spend 2-3 hours fixing specific, well-defined problems.
The hardest part about getting stuck at 80% isn't the technical problem...

It's watching your momentum die.

You go from "I'm building something amazing!" to "Maybe I'm not cut out for this" in about 48 hours.

That psychological shift? That's the real cost of the last 20%.
Problem: Developers were unsure about payouts

→ Solution: Improved payment explanation with clear descriptions and a simple flow.
Problem: Too many fields to define a task

→ Solution: Redefined the minimum data, trimmed fields, and added smart defaults and drafts facility for saving tasks to be added later.
Two weeks into our MVP. Here are some things we've done to improve in the last few days, thanks to the feedback of our users.

Problem: Too much friction to add a task

→ Solution: Moved “Add task” ahead of signup so users can create pre‑auth, then sign up to save.
5/ This isn't a failure of AI - it's just the reality of building real software.

The solution? Find developers who specialize in finishing what AI started.

#AITools #TechChallenges #Innovation #EdgeCases #ComplexSolutions
#Collaboration #Integration #AIChallenges #Debugging
3/ The last 20% is often about making things work together, not just making individual pieces work.

4/ AI can't debug production environments, handle deployment quirks, or troubleshoot server configurations.
Why do AI tools get you to 80% then leave you hanging? 🧵

1/ AI tools are trained on "happy path" scenarios - the common, straightforward implementations that work most of the time.

2/ Your specific use case probably has edge cases, integrations, or more complex requirements.
The outcome is pretty clear, but somehow we convince ourselves we should figure it out alone.

Why do we do this to ourselves?

#TimeManagement #AICosts #DeveloperEfficiency #ProductivityChallenges
Here's what that "simple fix" actually costs you:

Time spent: 6+ hours (you should be talking to users)
AI credits: £50+
Stress level: Maximum
Momentum lost: Priceless

Meanwhile, a developer could fix it in a fraction of time.
The hardest part about getting stuck at 80% isn't the technical problem...

It's watching your momentum die.

You go from "I'm building something amazing!" to "Maybe I'm not cut out for this" in about 48 hours.
But with no real technical understanding they can not describe the problems they are having.

If you are one of these people. Talk to us on our Discord.

We'll take the time to help you diagnose your issues in a way that gets your through your roadblock.
The big problem is that the first 80% goes really well for them. Then they want to redefine a few things, update some key functionality and all of a sudden they are burning credits at a crazy rate.

At this point they reach out for help.
I just had a really interesting support call with a customer.

They have an interesting issue which we are seeing more and more of.

Tools like Lovable and Base44 and empowering non technical people to try and create massive internal projects for bricks and mortar businesses.