Atlassian Cut 900 R&D, Kept 108 Grads: The 136-Person Forge Pool

Atlassian cut 1,600 roles on March 11, 2026, but the shape of the cut matters more than the number. More than 900 of those roles came out of R&D while the full 2026 graduate intake of 108 hires was preserved. That is a public statement about which skills the company thinks are scarce, and it points sourcers at a pool most of them are ignoring.

What Atlassian actually cut, and what it kept

Atlassian eliminated roughly 10% of its workforce on March 11, 2026, weighted toward experienced R&D and customer-facing roles, while keeping the February 2026 graduate class of 108 intact. The cut was framed as funding a shift to AI investment and enterprise sales, with restructuring charges of $225M to $236M booked to close by end of June 2026.

The relevant numbers, before you build a single search:

• 1,600 roles cut globally, about 10% of headcount
• 900+ of those roles came from R&D
• 108 graduates preserved in the February 2026 intake, on top of the 95 grads hired in February 2025
• Sydney absorbed roughly 35% of the cuts
• Severance ran $169M to $174M, office-space charges $56M to $62M
• Stock down ~33% for 2025 pre-selloff, more than 84% off the 2021 peak

Reporting from Kore1 suggests the top of the round was weighted toward sales managers and customer success leads, with engineering taking a smaller, more surgical share. That matters. It means the ex-R&D exits are quality-selected, not a broad dump of underperformers. A staff engineer walking out of Sydney in April 2026 was not the tail of the distribution.

The graduate signal

The loudest line in the whole announcement is the ratio. Atlassian cut 900+ experienced R&D roles while shipping 108 grads, an 8.3x ratio of senior-cut to junior-preserved. In October 2025 on 20VC, Mike Cannon-Brookes said tech creation is "not output-bound" and pledged more graduate hires in 2025 and 2026 than in prior years. Read together, this is leadership betting that the productivity-per-headcount curve now favors AI-native juniors trained on agentic workflows over senior distributed-systems ICs.

If your sourcing thesis is "grab senior or staff Atlassian alumni before HubSpot does," you are fishing in exactly the pond Atlassian just declared over-supplied. The interesting pool is narrower.

The Forge pool is 136 people. Globally.

In Refolk's index of professional profiles, only 136 people worldwide currently carry Forge plus Atlassian ecosystem skills in their profile. That is smaller than most Rust or Zig niches, and it is the actual addressable market for anyone trying to poach a Marketplace-native developer.

136

Top employers on that list are not who you would guess. Capgemini shows 3, Atlassian itself only 2, followed by Nagarro, Accenture, and Cisco. Concentration is India-heavy (Chennai, Bengaluru, Noida) with thin representation in Sydney, Melbourne, San Francisco, and the UK. This is a system-integrator-heavy pool, not a product-engineer pool, which changes the pitch you have to make.

The mechanism behind the scarcity is straightforward. Forge launched in 2021 as Atlassian's cloud app dev platform for Marketplace apps on Jira and Confluence, phased in behind the older Connect framework. The skill only accretes to engineers at Marketplace partners (DevSamurai, Automation Consultants, forge-apps.com, Hookshot) or inside Atlassian itself. Five years of adoption, one proprietary runtime, and now Rovo agents built on top of it. That is the entire supply chain.

Rovo Dev is a wedge, not a Copilot clone

Rovo Dev's distinctive strength is enterprise context, tying agent work back to Jira tickets, Bitbucket PRs, Confluence pages, and incident data. Competitors approximate this via MCP; Rovo does it natively. The target profile is not "another AI engineer." It is engineers with LLM tool-use experience plus Atlassian API depth, a much narrower Venn than either skill alone.

At Team '26, Atlassian opened the Teamwork Graph to external developers and AI agents via three channels, including Teamwork Graph Connectors going GA through Forge. Forge demand is expanding externally at exactly the moment Atlassian is shedding internal R&D. That is the arbitrage.

Where the ex-Atlassian talent pool actually is right now

Most of it is not in-market yet, because Australia, the EU, and the UK legally require consultation before role elimination. The 1,600 number was fixed on March 11, but exit dates roll out over months, so the ex-Atlassian talent pool is arriving in waves through Q2 and Q3 2026.

The Blind chatter this week ("Atlassian 10% layoffs confirmed", "CEO deeply sorry") is not new sentiment. It is a legal artifact. Sydney absorbed 35% of the cuts, and Sydney consultation windows are only now closing. If you started sourcing on March 12 and gave up on April 15, you missed the actual wave.

Here is a sourcing check on how invisible that wave still is:

The gap between "900+ R&D roles cut" and "4 self-labeled ex-Atlassian engineers currently searchable" is the whole game. Public "ex-Atlassian" self-labeling lags reporting by 3 to 6 months, especially where consultation law delays the actual last day. If you are sourcing off LinkedIn's "past company: Atlassian" filter, you are looking at 2024 exits, not the 2026 wave.

This is the exact gap Refolk is built for. You describe the person in plain English (Forge experience, Atlassian ecosystem, currently in Sydney or on the way out) and get a ranked shortlist that pulls from GitHub, the open web, and community forum activity, not just the LinkedIn field that ex-Atlassians have not touched yet.

The named free agents worth watching

Rajeev Rajan is the highest-profile free agent from the round. Atlassian's CTO stepped down March 31, 2026 after joining from Meta (VP Eng) with 20+ years at Microsoft before that. The role was split between Taroon Mandhana (CTO Teamwork, previously head of engineering for AI and products) and Vikram Rao (CTO Enterprise and Chief Trust Officer).

That split telegraphs where hiring is going, and it should shape your candidate lists:

• Mandhana's Teamwork org will pull agent, ML, and foundation-model talent. Source ex-Atlassian engineers with LLM tool-use, RAG, or agent-orchestration signals.
• Rao's Enterprise and Trust org will pull security, compliance, and platform reliability. Source ex-Atlassian engineers with SOC2, FedRAMP, IAM, and multi-tenant isolation signals.

The two orgs will fight for the same alumni. If you are placing candidates into HubSpot, Linear, Notion, or any Jira-adjacent buyer, you want to reach Mandhana's target profile before he does. Rao's profile is harder to poach because those engineers get retention offers first.

The gap between 900 R&D cut and 4 self-labeled ex-Atlassians searchable today is the whole game.

Where to actually look, this week

Four surfaces will get you further than any LinkedIn Recruiter search this quarter. Rank them by how well they beat the AU/EU/UK consultation lag.

1. community.developer.atlassian.com. Active thread authors on Forge and Rovo issues are self-identifying senior practitioners. Filter for authors who posted in the last 90 days on Forge runtime limits, Rovo Agent tool-use, or Teamwork Graph Connectors.
2. GitHub commits to Forge apps and Marketplace repos. The Atlassian org, the atlassian-labs org, and public Marketplace partner repos show commit history that predates any LinkedIn update. Contribution signal is real; job title is stale.
3. Marketplace partner rosters. DevSamurai, Automation Consultants, forge-apps.com, and Hookshot are the only off-Atlassian shops with real Forge plus Rovo depth. Their senior engineers are targetable and rarely on active recruiter radar.
4. System integrators (Capgemini, Nagarro, Accenture). The largest concentration of Forge-labeled developers per Refolk's index sits inside consultancies. The pitch is different (product ownership, equity, not just billable rate), but the pool is larger than the product-engineer pool.

If you want to shortcut all four surfaces at once, describe the profile to Refolk in plain English (something like "senior engineer with Forge experience, active in the Atlassian developer community in the last 6 months, based in Sydney or Bengaluru") and let it merge GitHub, LinkedIn, and open-web signals into one ranked list. The Forge pool is too small and too fragmented to source one platform at a time.

The 90-day window

You have roughly a quarter before this window closes. HubSpot, Linear, and Notion are shopping the same 136-person list, plus the delayed Sydney and European wave. Consultation windows will keep dropping candidates into the market through Q2 and Q3 2026, then the pool refreezes.

Three moves that pay off inside 90 days:

• Skip senior generalist ex-Atlassians. Atlassian just declared them over-supplied. The market will follow.
• Prioritize Forge plus LLM tool-use. That intersection is where Rovo Dev built its moat and where the next 12 months of Marketplace demand sits.
• Reach candidates before their LinkedIn updates. GitHub commit history and developer forum activity are 3 to 6 months ahead of the "past company" field.

The preserved graduate class was Atlassian telling you what it thinks is scarce. The 900 R&D cuts were Atlassian telling you what it thinks is not. Sourcing lists that ignore either signal will lose the next quarter.

FAQ

How many ex-Atlassian engineers are actually in-market from the March 2026 cuts?

Fewer than the headline suggests, because Australia, the EU, and the UK legally require consultation before role elimination. The 1,600 number was fixed on March 11, 2026, but exit dates roll out over months. Sydney alone absorbed 35% of the cuts and those consultation windows are only closing now. Expect the actual in-market wave to peak in Q2 and Q3 2026, not immediately after the announcement.

Why is the Forge developer pool so small?

Forge launched in 2021 as Atlassian's cloud app dev platform, phased in behind the older Connect framework, and the skill only accretes to engineers at Marketplace partners or inside Atlassian itself. In Refolk's index, only 136 professionals globally currently carry Forge plus Atlassian ecosystem skills, with concentration at Capgemini, Nagarro, Accenture, and Cisco. It is a proprietary runtime with a five-year adoption curve, which is why the pool is smaller than most Rust or Zig niches.

What does the preserved 2026 graduate intake actually signal?

That leadership sees the skills gap as junior AI-native engineers, not seasoned platform veterans. Atlassian cut 900+ experienced R&D roles while preserving 108 graduate hires (an 8.3x ratio), and Cannon-Brookes explicitly pledged more graduate hires in 2025 and 2026 than prior years. Sourcers running "senior staff Atlassian alum" plays are fishing in the pond Atlassian just declared over-supplied.

Who should sourcers target first from the ex-Atlassian talent pool?

Engineers with Forge or Rovo experience, then engineers with LLM tool-use plus Atlassian API depth, then security and compliance engineers from the Enterprise org. Rajeev Rajan is the highest-profile free agent, but the more useful targets are mid-to-senior Forge and Rovo engineers in Sydney, Melbourne, and Bengaluru whose LinkedIn profiles have not yet caught up with their exit dates. GitHub commits and Atlassian Developer Community forum activity are the leading indicators; the LinkedIn "past company" field is the lagging one.