distrust the accuracy of AI tools, and 33% trust it
More developers distrust AI tools' accuracy than trust it, in Stack Overflow's 2025 survey.
Locked in
70%
of AI-native developers feel locked into closed tools
Estimated
Most AI-native developers feel locked into closed tools, by our estimate.
Enterprise trust
60%
of enterprises name agent trust as a block to adoption
Most enterprises name trust in agents as a block to adopting them.
The open gap
USD 2bn
open agent marketplace, unclaimed
Estimated
None of the four tools we compared offers an open marketplace for agents. We estimate the gap at USD 2 billion.
Market growth
40%
a year, growth of the AI agent market
Projected
We project the market for AI agents to grow 40% a year.
What the analysis concluded
28
42
0100
28 · Difference from rivals42 · Market readinessThe product trails the market by 14 points. ForgeGrid would close that gap before scaling.
The first step
ForgeGrid seeds the agent marketplace before it opens.
The aim
The aim is to make the cost of closed tools visible. ForgeGrid's rival is the closed-tool model itself.
1 · The futures
Two unknowns decide which future comes. Seeding the marketplace wins in two, and stops the launch in two.
Will developers publish and sell agents on it? And can the coding tool match the leaders' speed?
A
Wins in two futures
ForgeGrid seeds the marketplace before it opens.
A funded builder programme, which pays developers to build agents, brings in 50 quality agents before launch, from developers who value owning their work.
B
Wins in two futures
ForgeGrid proves its speed first.
It publishes benchmarks against the leading tools, and checks speed on a developer's own daily work.
Neither question is answered yet. The builder programme answers the first, and published benchmarks answer the second. A and B are the two moves above: each card says what each would do in that future, and the sign to watch.
The tool keeps upThe tool lags behind
Builders publish and earnThe marketplace stays empty
01Look here first
The open alternative
Builders publish and earn, and the tool keeps up. Developers come for speed and stay for the marketplace.
A
ForgeGrid opens the marketplace once 50 agents are live.
B
ForgeGrid leads with the benchmarks, then shows the marketplace.
Sign to watch
A builder earns within a week of publishing.
02Look here first
A marketplace without the tool
Builders publish agents, but daily coding stays on closed tools. The marketplace and its audit trail carry the value.
A
ForgeGrid keeps seeding agents, and sells verified agent history to regulated teams.
B
ForgeGrid stops promising equal speed until the benchmarks pass.
Sign to watch
Teams ask for the audit trail before the editor.
03
One more fast tool
The tool keeps up, but few builders publish. It competes with closed tools on speed alone.
A
ForgeGrid holds the public marketplace back until it has agents.
B
ForgeGrid sells on speed and keeps the marketplace in beta.
Sign to watch
Beta developers stay, but publish nothing.
04
The cold start confirmed
Few builders publish and the tool lags. An empty marketplace would confirm the problem in public.
A
ForgeGrid does not open the marketplace.
B
ForgeGrid keeps the beta closed and spends nothing on growth.
Sign to watch
Fewer than 50 agents when the builder programme ends.
Where to look first
We would look at 01 and 02 first. Both keep the marketplace alive. The builder programme would show whether ForgeGrid is in that pair or in 03 or 04. Published benchmarks then separate 01 from 02.
An opening outside these futures
The EU AI Act sets transparency rules for some AI systems. If audits of AI agents become mandatory, verified agent history would matter in every future above.
2 · The gates
If ForgeGrid takes this road, three phases lead from approval with conditions to full approval, each with a measured output.
Phase 01 · Current phase
Seed and validate
ForgeGrid runs the builder programme, seeds 50 or more agents, matches the leaders' speed and runs a closed beta with 500 developers.
Gate 1
50+
Agents are live, with benchmarks and retention data.
If fewer than 50 agents are live when the builder programme ends, ForgeGrid keeps the beta closed and spends nothing on growth.
Phase 02
Prove the model
ForgeGrid opens the marketplace, measures transactions and tests pricing.
Gate 2
1,000
Builders are active, and revenue per developer is measured.
Phase 03
Scale on proof
ForgeGrid puts growth spend behind a platform that has proved itself.
The recommendation
Full approval
A scale commitment you can defend.
The three gaps the phases close
The marketplace starts empty.
Without enough agents, the platform has nothing to offer.
The tool trails the leaders on speed.
Developers would not accept friction for ideals.
The revenue model is unproven.
Fees on transactions are small until the marketplace is big.
What to avoid
A marketplace with no agents is worse than no marketplace at all.
3 · The tension
The market window is open, and the product is what holds it back.
The opening
An estimated 70% of AI-native developers feel locked into closed tools.
What holds it back
Developers would not accept a slower tool for the sake of openness.
The opening
None of the four tools we compared offers an open marketplace for agents, a gap we estimate at USD 2 billion.
What holds it back
The marketplace starts with no agents in it.
The opening
One platform serves both developers and agent builders.
What holds it back
Web3 sign-up, which uses blockchain tools, turns mainstream developers away.
4 · The evidence
The analysis scored five domains. Four bear on whether to scale now, and each reads cautious.
5Domains scored
→
28Factors analysed
→
141Markers
→
4Futures mapped
The outcomeApprove with conditionsIt would reach approve once the marketplace shows real transactions.
DomainReadingScoreLifted byHeld back by
MarketCautious47
A new category, unclaimed and growing.
The market leader's large user base.
ProductCautious47
Its design keeps the cloud's speed and gives developers ownership.
An empty marketplace, and no proven speed.
ExperienceCautious44
One platform for developers and agent builders.
Sign-up friction for mainstream developers.
CustomerCautious42
Clear demand for open alternatives.
Only 33% trust AI output.
What the market shows
Four alternatives cost USD 10 to 20 a month. All are closed.
None offers an agent marketplace or verified reputation.
ForgeGrid's business case shows a 98% gross margin. It borrows a software-business model and has not been checked against a marketplace.
The analysis puts the base-case return below the hurdle, so the economics support the validation phase, not aggressive growth.
5 · Who says yes
Three buyers weigh the same tools, and each needs a different proof.
First proof
The pragmatic accelerator
Senior developer or tech lead
Wants
Getting back time lost to switching tools.
Says yes when
Benchmarks are run on their own daily work, with no ideology in them.
Second proof
The agent builder
Agent developer or open-source contributor
Wants
Building and selling agents without lock-in.
Says yes when
The marketplace works, and they can publish an agent and earn within a week.
Last proof
The sceptical verifier
Engineering manager or platform lead
Wants
Agent quality that can be checked.
Says yes when
Each agent has an audit trail and an outside security review.
6 · What it cannot answer
The USD 2 billion gap is an estimate.
It is drawn from neighbouring AI software markets, so it is only a rough size.
The developer data comes from general surveys.
None of it comes from ForgeGrid's own users, so friction is estimated and not measured.
The first phase has no budget yet.
The cost and the time needed to fill the marketplace are still open. The builder programme would give the first figures.
Sources: Stack Overflow Developer Survey, 2024 and 2025 · Arcade.dev developer statistics · the platform's own documentation · market research. The competitors are counted, not named.
We live in a decision economy. Companies now compete on how well they choose where the money goes.Most industries have changed how they work, but we still decide in slides. A slide is static and slow to build, and it falls behind when the market moves fast.In five years, we see strategic decisions as comparable as stocks on a Bloomberg Terminal. Every team puts its decision on the same footing. The people who evaluate them can see them as they change, and act sooner.