Samosky Dev (@samsonajulor) header
Samosky Dev (@samsonajulor) profile picture

Intern @quikdb_online — rebuilding the cloud as a decentralized network. Helping devs own their data, scale globally, and deploy with freedom.

Qatar
Joined October 2017
Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)Photo by Samosky Dev (@samsonajulor)
Proud to share that @quikdb_online is one of the 8 teams selected for the final stage of the @ayahq_ x @LiskAfrica @LiskHQ Incubation Program! Big thanks to the mentors and team behind this — the support has been a game changer. Let’s keep building. 💪
They came in with big ideas. Now, they’re building real solutions. After weeks of refining, testing, and mentorship, 8 promising startups have advanced to the final stage of the @ayahq_ x @LiskAfrica @LiskHQ Incubation Program, the 2 week Founder Summit, happening August 18–30 at the AyaHQ x Lisk Builder Hub Accra, Ghana. These teams didn’t just show up with prototypes, they showed up with purpose. Each one is tackling a real need in Africa’s digital and economic landscape, and they’ve made incredible progress in just a few weeks. We’re bullish on these projects because they’re building scalable, user-driven tools that address critical challenges: → Financial inclusion through crypto payments and stablecoin rails → Treasury tools and programmable wallets for onchain businesses → Platforms for creators, freelancers, and event organizers → Decentralized infrastructure that prioritizes data ownership and usability Each project stood out for: → Strong technical execution → Clear value propositions → User-first design → Real-world momentum
5
8
40
2,921
A running instance can still contain an unhealthy accelerator. Reliable recovery starts with the right signal, replaces the impaired capacity, and verifies that the workload can run again.
7
A pod boundary is a design choice, not a security conclusion. If code is untrusted, multi-tenant or privileged, ask whether sharing the host kernel matches the actual risk. Start with the workload’s trust level, then choose the boundary.
1
18
Finding a leaked secret is not the same as containing it. The response still needs an owner, revocation, replacement and a check for where the value was used. A good alert shortens discovery. A good runbook shortens exposure.
1
1
23
Logs are part of the security boundary. A token can avoid the repository and still leak through a build command, error message or debug step. Review what your pipelines print, who can read it and how long it remains available.
17
CI: “I only ran npm install from the fork with production secrets available.” Security reviewer: “That is a fascinating use of the word ‘only.’” Trusted workflow. Untrusted code. Very bold pairing.
18
The demo was the easy part. Production asks the slower questions: identity, cost, observation, recovery and failure boundaries. The model is visible. The operating system around it is the product too.
15
Inspect is not execute. A privileged CI job can safely read a contributor's PR as data. The boundary breaks when a step runs that fork's script or dependency install with secrets in reach. Audit the actual commands not just the event trigger.
19
A friendly reminder: Token expiry is NOT the same as token rotation. Yes, a short-lived token is great for limiting exposure. But without automatic refresh, you're just turning a "security win" into a daily task users hate Make the safer path the easier path.
17
Not everything needs a lock. Immutable work can spread out. The small shared reference still needs agreement. Before adding coordination, ask: what exact piece of truth can two writers change at once?
17
If only every secret wore a name tag: HELLO, I AM A PRODUCTION CREDENTIAL. Until then, “what can this value authorize?” is a better review question than “does it look like a key?”
26
A secret is not risky because it starts with a familiar prefix . It is risky because of what it authorizes. Patterns catch what happened. Context explains why it matters.
20
“It does not look like a key.” That is not the safety test. The better question is: what can this value authorize? Pattern checks help but Context checks catch a different class of risk.
16
"it works in staging, but fails in production.” Same container, same manifests yet different behavior. Why? Because the runtime contract changed across networking, storage, resources or privileges. That contract is what the new Workload Conformance group is making testable.
1
2
30
When an agent proposes, approves and executes its own action, the approval is just decoration. Real control requires an external, deterministic control.
1
1
32
"The prompt says it cannot deploy.” Okay. Does the credential agree? A prompt is an aspiration. IAM and blast radius are reality. If an agent’s guardrails rely entirely on instructions it can read, evaluate or ignore you don't have security. You have wishful thinking.
1
19
The bug was one line. The environment was the wrong one. Suddenly the “quick fix” has a meeting invite.
15
A rollback plan that begins after failure is not a rollback plan. Before production, define the trigger, the owner and what happens to new data. “Go back” is not enough when the system has already changed state.
1
16
The service answered. A small health endpoint can show that the service responds. A separate user-path test shows whether the product can complete the important job. probe the service, test the product.
16
Your homepage is not a health check. Use a dedicated, minimal endpoint for the operational signal. Then test the real user path separately. One proves the service can answer. The other proves the product can work.
17
The safest deployment credential is the one that expires. For CI/CD, prefer short-lived, scoped access where your cloud and workflow support it. Then review the trust rule. Temporary access with a loose boundary is still a loose boundary.
18