Anthropic Took Part of Opaque. The Leftover Pool Is 9 Deep.

On Oct 13, 2026, Anthropic announced a licensing plus acquihire deal with Opaque Systems, taking the confidential-AI tech and only part of the team. CEO Aaron Fulkerson is going. CTO Rishabh Poddar is likely going. The engineers Anthropic passed on are, for about 72 hours, the most sourceable confidential-computing talent on the open market.

Then Google, Microsoft, Nvidia, and the sovereign-AI buyers that already partnered with Opaque will close the window. If you source for a bank, a cloud, a sovereign lab, or any enterprise running agents on trusted execution environments, this is the week to move.

The leftover pool is 9 people, not the full team

The realistic sourceable ceiling from this acquihire is 9 current Opaque Systems engineers. That is what Refolk's index shows as the number of profiles currently employed at Opaque, and it is the number that matters because Anthropic took "some" engineers, not all, and the executives were the first to go.

Zoom out and the market context gets sharper. Opaque's technology began as a UC Berkeley RISELab project about 10 years ago, carried the company to a $300M valuation after a $24M Series B in February, and attracted $55.5M in total funding. The engineering output per dollar is high and the headcount is small. That is the whole point of a confidential-computing startup: a handful of people who understand enclave runtimes, attestation pipelines, and MC² internals well enough to ship them to banks.

4.7%

The reason sourcers need to move in days, not weeks, is the denominator. There are 417 profiles globally with "Confidential Computing" as a listed skill in Refolk's index. The US holds roughly 46% of that, 191 profiles. Opaque's 9 current engineers are 4.7% of the entire US pool. There is no comparable alternative cluster you can tap next quarter. Fortanix, Edgeless Systems, Phala, and Super Protocol are each similarly thin, and poaching from Intel's own SGX group is a slow exercise.

The numbers behind the window

Here is the table-ready version of the pool, pulled from Refolk's index of professional profiles.

Two numbers in that table do most of the work. The 80 is your outreach universe for anyone who ever touched the stack, including people who left years ago and are now at Google or Microsoft and quietly interested in the next shop. The 9 is this week's exclusive: current employees who did not get an Anthropic offer, who are updating their LinkedIn right now, and who will not be on the market by Friday.

Top titles among the Opaque Systems profiles in Refolk's index skew IC and senior: Senior Software Engineer (2), Principal Engineer, Backend Software Engineer, Chief Platform Officer, and CTO & Co-Founder. That is a feature, not a problem. Fulkerson and Poddar going to Anthropic is a sourcing signal, because the ICs who built the enclave runtime, the attestation pipeline, and the open-source MC² libraries are exactly the hires enterprise buyers actually need. Not the executives.

Who the leftover engineers are

The leftover pool is dominated by backend ICs with hands-on TEE and Intel SGX experience, not by executives or go-to-market staff. Anthropic took the brand names. The code was written by the people still at their desks.

Names and anchors from the Opaque engineering graph worth searching for:

• Aaron Fulkerson, CEO, confirmed going to Anthropic. Exclude him.
• Rishabh Poddar, CTO and co-founder, likely going to Anthropic. Treat as taken.
• Raluca Ada Popa, co-founder, Berkeley faculty, not moving. She anchors the alumni network you want to tap for warm intros.
• Ion Stoica, co-founder, Berkeley faculty and Databricks chairman, not moving. Same role: a reference, not a hire.
• MC² project contributors on GitHub, especially those with Berkeley .edu addresses who never joined full-time but shipped meaningful commits.

The last bullet is the one most sourcers will miss. The MC² codebase, the RISELab and Sky Computing Lab lineage, and the Open Confidential Computing Conference (OC3) speaker lists together form a Berkeley talent graph that extends well past Opaque's payroll. GitHub contributor history on the mc2-project org is a better signal here than LinkedIn, because the people who shipped the hard parts often have 4-line LinkedIn profiles and 400-commit GitHub histories.

You do not have time this week to assemble a Boolean string, run it across three sources, dedupe, and chase down GitHub handles. Describe the person in plain English, including "ex-Opaque, not already at Anthropic," and get the ranked list back.

Who else is already dialing them

Expect Google, Microsoft, Nvidia, and Pattern Labs to be in the inbox by the end of the week, with sovereign-AI buyers close behind. Opaque developed its verifiable AI tools in collaboration with Nvidia, Intel, Microsoft, and Google, which means those teams already know the names, already have internal champions, and already have reqs open.

The likely competing buyers, ranked by how fast they will move:

1. Google, confidential GKE and confidential VMs team. Prior Opaque partner, deep Bay Area footprint, fastest internal referral path.
2. Microsoft, Azure Confidential Computing. Prior partner, broader TEE product surface, slightly slower internal process.
3. Nvidia, Hopper and Blackwell confidential compute. Prior partner, hardware-side demand, highest comp ceiling.
4. Pattern Labs, Anthropic's June 2025 collaborator on confidential inference. Small, fast, credible adjacent destination.
5. Fortanix, Edgeless Systems, Phala Network, Super Protocol, Secretarium. Pure-play confidential-computing shops. Smaller comp bands, but the engineers care about the problem.

The non-obvious angle is the sixth buyer group: Opaque's own former customers. Neither Anthropic nor Opaque has confirmed whether the platform keeps serving ServiceNow, Accenture, and the rest under the license-and-hire structure. Those customers now have both motive (support risk on an in-production confidential-AI platform) and budget to directly hire leftover Opaque engineers for in-house teams.

Pitch leftovers to Opaque's former customers before you pitch them to competitors. The customer-side offers close faster, the comp is often better than a lateral to Google Cloud, and the engineer gets to keep working on the same code they already wrote.

Pitch leftovers to Opaque's former customers before you pitch them to competitors. The customer-side offers close faster.

The 72-hour sourcing playbook

Run three parallel tracks in the first three days, not sequentially. The pool is small enough that whoever reaches each engineer first wins.

Track 1: The current-9

Direct outreach to the 9 current Opaque Systems employees in the index. Reference the Oct 13 announcement explicitly. Do not pretend you do not know what happened. Lead with the specific team you are hiring for and the specific piece of their work (enclave runtime, attestation, MC² libraries, SGX integration) that matches the role. Generic "exciting opportunity" messages will lose to the Google recruiter who names the actual GitHub PR.

Track 2: The ex-Opaque 71

Of the 80 "Opaque Systems" profiles in Refolk's index, 9 are current. The other 71 are alumni, scattered across Google, Microsoft, Intel, and smaller TEE startups. Many of them are quietly watching the news and wondering if their old colleagues got a better deal. This is a warm market for the next 10 days. Message them before the news cycle moves on.

Track 3: The Berkeley graph

Pull the mc2-project GitHub org contributor list. Cross-reference with Berkeley RISELab, Sky Computing Lab, and recent OC3 speaker rosters. You are looking for grad students and postdocs who shipped Opaque code without ever joining the payroll. These people are not on any competitor's radar yet, which is the whole advantage.

Geographic arbitrage: the EU and UAE angle

Opaque's expansion into Europe and the UAE means some leftover engineers may be remote or EU-based and sourceable at well below Bay Area comp. Global confidential-computing talent is thinly spread across Zurich, London, Stockholm, Brussels, Munich, Delhi, and Rome, which means an engineer in Zurich or Munich faces almost no local competition for their specific skill set and is often reachable by sovereign-AI buyers like Mistral, Aleph Alpha, and G42 at a comp number a Bay Area lab would laugh at.

If you source for a sovereign lab, this is the week to run a parallel EU search. The 46% US share leaves 54% of the global pool everywhere else, and the Opaque EU footprint means some of those 54% have shipped against the same codebase as the SF-based engineers Google is about to overbid for.

What the pool looks like next week

By roughly Oct 20, the window closes and the 9 become 2 or 3. The pattern on acquihire leftovers is consistent: within a week, the strongest ICs are already in final rounds with the obvious buyers, and within two weeks, offers are signed. The alumni-71 stay sourceable, but the exclusive signal of "just-displaced, not-yet-placed" is gone.

The 417 global confidential-computing engineers do not grow meaningfully quarter over quarter either. The skill takes years to build, the hardware changes slowly, and the people who do it well tend to stay in it for a decade. Which is why losing 9 to a single acquihire feels, from the market's perspective, like losing a non-trivial percentage of the entire pool. Because it is.

Source the Opaque leftovers this week. The next window on confidential-computing talent at this concentration will not open until the next acquihire, and the next acquihire may well be the one that leaves nothing behind.

FAQ

How many engineers did Anthropic actually take from Opaque Systems?

Terms were undisclosed. Anthropic confirmed it licensed Opaque's technology and hired "some" engineers, with CEO Aaron Fulkerson confirmed as joining and CTO Rishabh Poddar likely joining. Refolk's index shows 9 current Opaque Systems employees, which is the realistic ceiling of the leftover pool. The remaining 71 "Opaque Systems" profiles are alumni already placed at other employers, mostly Google, Microsoft, and Intel.

Why do I only have 72 hours to source these engineers?

Because the US confidential-computing pool is 191 engineers in Refolk's index and the Opaque leftovers are 4.7% of it, which means Google, Microsoft, and Nvidia (all prior Opaque partners) already know the names and have internal champions ready to push reqs through. Historical pattern on acquihires: within a week the strongest ICs are in final rounds, and within two weeks the offers are signed. The exclusive "just-displaced" signal is gone by Oct 20.

Should I pitch Opaque leftovers to competitors or to Opaque's former customers?

Pitch customers first. ServiceNow, Accenture, and other enterprise Opaque customers now face platform-support uncertainty and have both motive and budget to hire in-house confidential-AI teams. Customer-side offers often close faster than lateral moves to hyperscalers, comp can be competitive, and the engineer gets to keep working on the same code. Pitch Google, Microsoft, and Nvidia second, knowing they will likely outbid you if the engineer wants a pure research seat.

What signals matter more than LinkedIn for this pool?

GitHub contribution history on the mc2-project org, Open Confidential Computing Conference (OC3) speaker lists, and Berkeley RISELab or Sky Computing Lab alumni rosters. The hard parts of the Opaque stack (enclave runtime, attestation pipeline, SGX integration) were often shipped by people with sparse LinkedIn profiles and heavy GitHub histories. If you only search LinkedIn, you miss the Berkeley grad students and postdocs who are not on any competitor's radar yet.