What changed

The navigation is four links: Now, Work, Journal, About. Search is a utility. The Archive, Publications, and CV are still here, one click down.

Clio Coder is on the front page because it is the thing I am proudest of right now. CLIO and IOWarp explain what it belongs to and who builds the rest.

The project atlas accounts for every repository I own, public and private, grouped by what it is rather than dumped as cards. Private prototypes are named with a one-line summary and no link. Four repositories are deliberately left out: personal machine configuration, home-lab configuration, personal finance tooling, and a retired website prototype. Nothing else is hidden.

The Journal is one stream. Essays, tutorials, field notes, updates, and the occasional personal piece are filters, not sections. The three 2024 reflections are here unchanged, with their old /blog/ URLs redirecting; the two 2024 updates were rewritten with links to what the projects became.

What did not change

Every old public URL redirects: /projects/, /blog/, /teaching/, /service/, /talks/, /news/, the individual project and post pages, and all the publication PDFs. Individual publication pages keep their citation-key URLs. If you find a broken link, tell me.

How it is built

Jekyll, Liquid, semantic HTML, one SCSS file, and a small dependency-free script. No framework, no analytics, no trackers. Content lives in validated data files and Markdown; a validator refuses to build if a fact lacks a source, a private URL leaks, or a repository quietly drops out of the atlas. Static redirect pages are generated from one manifest and checked in CI.

Next: more short notes, closer to the moment the detail is still warm.

Prerequisites

• Node.js 22.19 or newer and npm.
• Linux or macOS. Windows is best effort until a stable release.
• A local inference server. Any of llama.cpp, Ollama, LM Studio, vLLM, or SGLang works; the runtime ids are llamacpp, ollama-native, lmstudio, vllm, and sglang.
• A model loaded in that server. If you have one GPU with 24 GB or more, the release was hardened against Qwen3.8-27B (the unsloth/Qwen3.8-27B-GGUF quantizations) on llama.cpp and LM Studio. A 4-bit quantization at 131072 context fits in 24 GB with a q8_0 KV cache.

Version as of August 26, 2026:

$ clio-coder --version
Clio Coder 0.3.7

1. Install

npm install -g @iowarp/clio-coder
clio-coder --version

From source instead, pinned to the release:

git clone --branch v0.3.7 https://github.com/iowarp/clio-coder.git
cd clio-coder
npm run install:local
export PATH="$HOME/.local/bin:$PATH"
hash -r

npm run install:local builds the CLI, links it at ${CLIO_CODER_BIN_DIR:-$HOME/.local/bin}/clio-coder, and initializes the home directory. If an older install is on your PATH, command -v clio-coder shows which binary the bare name reaches.

2. Start the model server

For llama.cpp, start it with the flags the README asks for so tool calls and reasoning parse cleanly:

llama-server --jinja --reasoning on -m /path/to/Qwen3.8-27B-Q4_K_M.gguf -c 131072

LM Studio needs nothing beyond loading the model. Note the port your server listens on; 8080 is llama.cpp’s default and 1234 is LM Studio’s.

3. Configure a target

The interactive way:

clio-coder configure

The wizard lists runtimes, asks for the endpoint and model, probes it, and saves it as both the chat and worker target. Bare clio-coder opens the same wizard when nothing usable is configured.

The scripted way, which does the same thing without prompts:

clio-coder configure --id local-lmstudio --runtime lmstudio \
  --url http://localhost:1234 --model your-model-id \
  --set-orchestrator --set-fleet-default
clio-coder targets --probe

clio-coder targets prints every configured target with its tier, runtime, auth state, URL, model, health, and capability flags. --probe actually connects. The model id has to match what the server advertises: if LM Studio has a different model resident than the one in your target, the turn fails immediately with a message naming the resident instances instead of silently routing to whatever is loaded. That is deliberate.

4. Run one headless turn

From any repository:

clio-coder run "Summarize this repository layout and its entry points." --json

--json streams JSONL events: a session record with the session id, working directory, target, model, and Clio version; agent_start, turn_start, message_start, message_end, turn_end, and agent_end; and tool events in between when the model calls tools. The agent_end event carries measured usage (input, output, cache reads and writes, reasoning tokens, API calls, and cost in dollars). Add --autonomy read-only if you want the guarantee that nothing can be mutated or executed on this turn; every mutating call is denied and the denial is recorded.

To hand the same task to a fleet agent instead of the main loop:

clio-coder run "<task>" --agent coder

That run gets its own receipt.

5. Read the receipt

clio-coder evidence list
clio-coder evidence inspect <runId>

A receipt covers routing intent, the resolved route, worker attestation, priced cost, phase timing, tool activity, safety decisions, and result conformance. Inside an interactive session, /view verify <runId> shows the same thing. clio-coder trace reads the durable dispatch trace.

6. Go interactive

cd your-project
clio-coder context init     # drafts CLIO-CODER.md from the source tree
clio-coder

context init writes the project handbook Clio loads every session and can adopt an existing CLAUDE.md, AGENTS.md, or GEMINI.md with provenance. In the session, /settings changes the model or autonomy level, /context and /cost show what is in the window and what it costs, and /help lists the rest.

Limitations

• Clio Coder is experimental. Interfaces may change between minor versions, and model-specific behavior varies by target. Keep the repository under version control and review what it proposes.
• A local model’s usefulness depends on the quantization, the context length you can afford, and whether tool calls parse cleanly. The model catalog in the repository records what has been checked per family; other models may need --jinja or a different chat template.
• The receipt records what the harness observed. It does not make the model’s reasoning correct.
• Connecting a Claude or ChatGPT subscription over OAuth uses the same path as the vendors’ own tools; using subscription credentials outside a vendor’s first-party apps may not align with their terms. That is your call, not Clio’s.
• Windows is best effort.

If something in here does not match what you see, clio-coder doctor plus clio-coder --version and node --version is what I need in a bug report. Redact secrets and private prompts first.

None of those are prompt-writing failures. They are systems failures, and I have spent the last two years building around that observation.

The thesis

CLIO gives agents what databases give applications: structured, reliable, queryable state. A model on its own has a context window. A scientist has terabytes on a parallel file system, a Slurm queue, instrument metadata, a decade of provenance, and tools with side effects. Something has to select from that world, shape it for one decision, and record what it chose. If that something is invisible, the agent’s answer is unauditable no matter how good the model is.

So “context” in CLIO is not a paragraph pasted before a question. It is a runtime product: content, metadata, current state, provenance, tool contracts, policy, and the boundary of human approval, assembled for a particular action and written down.

What that looks like in code

Clio Coder is the component where I have pushed this the furthest, because I write all of it.

Selection is bounded. Every observation tool returns through a truncation envelope with offload paths and next hints, so a large result is bounded rather than silently cut. One grep cannot blow the window. The compiled prompt and tool schemas stay byte-stable across turns so a local model’s prefix cache stays hot, and the ledger records a cache verdict per call.

Authority is typed. Twenty tools sit in seven policy planes. Bash is default-deny. The autonomy dial decides what runs immediately and what parks for approval; it never widens the tool surface. A worker can only narrow the orchestrator’s authority, never exceed it. Reviewers and judges run read-only. Since v0.3.7 a dispatch carries typed intent: read roots, write roots, expected outputs, and the verification checks the host runs afterwards with no shell.

Everything leaves a receipt. A run seals tokens, priced cost, tool activity, safety decisions, routing, worker attestation, and result conformance. clio-coder evidence inspect reads it back. Nothing in the audit trail is reconstructed from prose, because prose is exactly the thing you cannot trust an agent about.

That is the shape I want across CLIO, not only in the coder: selection, rejection, authority, and action, with a receipt another person can check.

More context is not the goal

Context windows invite an easy metric: fit more tokens. Scientific context does not become correct when it becomes large. A smaller, well-scoped package with provenance beats a giant retrieval dump. A refusal caused by missing authority beats a plausible action. An explicit conflict between two measurements beats a smooth synthesis that hides the disagreement.

The objective is the smallest sufficient context for a bounded action, with enough provenance and state to make the result inspectable.

Context is a data-movement problem

At scientific scale the right evidence lives in object storage, a parallel file system, an instrument buffer, or another site. Moving all of it toward a model wastes bandwidth; summarizing too early erases the feature the agent needs. Where filtering happens, which representation moves, what gets cached, when a derived object is safer than the source: these are the questions Hermes and LABIOS asked about I/O, and IOWarp’s transfer and assimilation engines ask them again with an agent on the other end.

What is unresolved

I do not have a universal schema for context and I doubt one exists across instruments, simulations, and organizations. Provenance describes a chain without proving its inputs were right. A perfect receipt does not repair a biased selection policy. Retrieval quality, adversarial content in retrieved text, changing permissions, and the cost of keeping history are all open. And context can expose the thing a system is meant to protect, so minimization and redaction have to happen before data reaches a model, not after generation.

The model matters. But the model acts inside the reality the surrounding system constructs. Design, test, and audit that layer like any other component.

Reflecting on 2024, I am proud to share a year filled with remarkable accomplishments. One of the most significant milestones was helping secure the five-year NSF award that gave IOWarp room to grow. The project was a culmination of years of dedication and passion for my field. Building it with a diverse team across universities, companies, and national laboratories was not just a professional milestone but a testament to the collaborative spirit and innovative potential of our community.

Beyond IOWarp, my year was marked by significant milestones: nine published papers, new PhD students joining my team, and two senior students successfully proposing their dissertations. A particular highlight was my transition from conference participant to organizer. This new role as part of the organization committee was a first for me, fostering my growth as a community leader. It was an exciting and rewarding experience, offering fresh perspectives and enhancing my contribution to the field in ways I hadn’t anticipated. This evolution in my professional journey was both challenging and deeply satisfying, underscoring a year of multifaceted growth and achievement.

Lessons from Challenges

However, 2024 wasn’t without its challenges. Professionally, I encountered moments where my contributions felt undervalued. This led to tough conversations and moments of self-doubt. But through these challenges, I learned a crucial lesson: value and appreciation should come from within¹. It’s easy to fall into the trap of seeking external validation, but true empowerment comes from recognizing one’s own worth².

On a personal level, ending a two-year relationship was a significant turning point. This experience taught me about the importance of self-worth and the courage needed to make difficult decisions³. It was a reminder that personal well-being is crucial and that sometimes, ending a chapter is necessary for growth. This experience also highlighted the power of introspection and self-reflection, leading me to a deeper understanding of who I am and what drives me.

Authenticity: The Key to Growth

A central theme of this year has been the importance of authenticity⁴. Whether in personal relationships or professional roles, being true to oneself fosters genuine connections and growth. For many years, I found myself conforming to societal roles, losing touch with my authentic self. This year, I embraced the power of being genuine, realizing that authenticity is not just about self-expression but also about finding those who appreciate you for who you truly are⁵. This shift in mindset has been transformative, fostering more meaningful connections both personally and professionally.

AI: A Companion in Self-Discovery

Looking ahead, I’m eager to contribute to the development of AI in a way that democratizes its benefits. AI should not just be a tool for the privileged few; it has the potential to bridge gaps in access to knowledge, productivity, and creativity for everyone. My focus on agentic AI and smart agents in data management aims to create systems that interact with humans in a natural, intuitive way, making data more accessible and actionable.

I see AI as more than a tool—it’s a partner that learns, adapts, and evolves alongside us. It’s a companion that empowers us to focus on what truly matters: creativity, problem-solving, and human connection. As society continues to shape AI, we also have a responsibility to guide its evolution ethically and inclusively, ensuring it reflects our collective values.

Reflections and Resolutions

2024 has been a year of profound growth, both personally and professionally. I’ve learned to value authenticity, embrace challenges as opportunities for growth, and look inward for validation rather than relying on external recognition. These lessons have not only shaped who I am today but have also set the tone for what I want to achieve moving forward.

As 2025 approaches, I’m excited to build on the foundation laid this year. I’ll continue to refine my vision for agentic AI, develop new projects, and explore opportunities for growth and progress. Personally, I aim to be more socially engaged, foster meaningful connections, and embrace new experiences that align with my authentic self.

Closing Thoughts

If I were to sum up 2024 in one word, it would be “transformation.” It’s been a year of change, introspection, and empowerment—a year that has prepared me for the challenges and opportunities ahead⁶. As I step into 2025, I’m filled with hope and excitement for the future, ready to embrace new adventures, create meaningful impact, and continue evolving in ways that matter most.

To those who’ve been part of my journey this year—thank you. You know who you are and I couldn’t be more proud of my connection with you⁷.

Here’s to a bright and fulfilling year ahead for all of us.

Some resources

Why Does Change Feel So Painful?

Change hurts because it disrupts our rhythm. We’re wired to crave stability and familiarity. When circumstances shift, our instinct is to hold on to what we know, focusing on the loss: the comforting routine we once followed, the partner with whom we shared late-night conversations, the job title that gave us our sense of identity. In my own life, I’ve felt this pain acutely. Serving in a disciplined environment like the military and then moving into the more fluid world of academia has taught me just how deeply our expectations and surroundings define us. When those expectations vanish, we’re left exposed and uncertain.

When approached with curiosity and resilience, these moments of unease can lead to transformative insights and achievements.

But does it have to hurt this much? I’ve found that pain is not a requirement, but a symptom of how we frame the experience. When I look back, I recall moments of personal upheaval—times when I had to leave behind communities, comrades, and even personal comforts. The ache was real, but I also discovered a startling truth: each ending cleared space for something new and meaningful to emerge.

Shifting the Narrative: Infinite Possibilities

Instead of dwelling on what’s lost, we can shift focus to what stands to be gained. Consider this: if you’ve recently parted ways with someone important, it’s natural to miss the routines you shared—your favorite show watched side-by-side, the familiar bar where you both unwound. But that same absence creates room for new interests, new communities, and new relationships. When I once faced a significant personal shift, I channeled my energy into exploring new technologies—immersing myself in learning about Large Language Models (LLMs) and cutting-edge software. By diving into something fresh, I not only distracted myself from the sting of loss but opened doors to professional opportunities, potential friendships, and yes, even the possibility of meeting someone who shares my passions.

It’s not just technology or hobbies. Perhaps you’ve never considered taking up a fitness class, joining a local volunteer group, or attending industry conferences that might broaden your horizons. Each leap into unfamiliar territory, each embrace of a new skill or social circle, transforms that initial pain into a platform for reinvention. Whether you choose to embrace these changes or not, the experience is transformative. It’s a chance to learn, grow, and make new connections.

Balancing Adaptability and Commitment

Of course, there’s a balance to be struck. I’ve learned from my time mentoring both soldiers and students that adaptability is a superpower. The principle of “quit often and quit early,” for example, encourages us to exit situations that don’t serve our well-being. It opens the door to alternatives—new jobs, new friendships, new lifestyles—that might fulfill us more deeply.

However, too much quitting can lead to shallow connections. We need to ensure that while we remain open to new possibilities, we also invest time and energy in cultivating depth. It might mean choosing to stick with a challenging relationship or a difficult project a little longer, not out of fear of change, but out of respect for what genuine bonds and long-term goals can bring.

Grounding in Past Successes and Global Perspective

When facing a new wave of change, remember the transitions you’ve already overcome. Every milestone in your life’s journey—whether it was surviving the intensity of basic training in the military, adapting to the freedom of college life, or finding your footing in a new city—proves that you’re not defined by static circumstances. You’re defined by your ability to evolve.

Grounding yourself in your personal history can also help you maintain perspective. I often remind myself that life is a series of waves. Some crests carry us high, others dip low, but the motion is constant. Recognize that change is universal; it’s happening in everyone’s life, all the time. You’re part of a global tapestry where millions of others are also trying to find their equilibrium amid shifting sands. Connecting with others who share their own stories—whether through local support groups, online forums, or even professional networks—can reinforce that you’re not alone.

Embracing the Duality of Change

Change carries a dual nature: it can feel catastrophic and liberating, punishing and rewarding, fearful and exciting. I’ve had the privilege of seeing thousands of individuals from diverse backgrounds face life-altering transitions. Across socioeconomic statuses, races, genders, and identities, I’ve observed that what differentiates those who flourish from those who stagnate is the willingness to embrace the unknown and reframe adversity as an opening rather than a closing.

From my own perspective—having lived multiple “lives” across military, academic, and communal environments—I know that I’m not immune to the pain of change. But I’ve learned to interpret that pain differently. Instead of treating it as a sign that something’s wrong, I see it as a cue that I’m about to learn something new about myself. This reframing doesn’t numb the discomfort entirely, but it converts fear into a sense of adventure, inviting me to ask, “What’s next?”

Harnessing Change for Growth

As someone who has guided thousands of soldiers, mentored countless students, and navigated my own shifting tides, I can attest: change will keep coming. It’s an inevitable constant in life, touching everything from our careers and relationships to our personal passions. But by reframing how we see it—from focusing on what’s lost to celebrating what could be—change can cease to be a source of paralyzing pain and become a catalyst for becoming someone stronger, wiser, and more resilient.

Take small steps toward this mindset shift. Try a new hobby, engage in a new network, or read up on success stories from others who’ve transformed their lives after major setbacks. Remember that you hold the power to define what change means to you.

Change hurts—but it doesn’t have to! It can heal, expand, and elevate you, if you let it.

Some Resources

• Positive thinking: Stop negative self-talk to reduce stress. (Mayo Clinic, n.d.) Mayo Clinic
• How to stop languishing and start finding flow. (Grant, A., 2023) TED
• Resilience. (American Psychological Association, n.d.) APA
• Stress. (American Psychological Association, n.d.) APA
• How to Let In New Perspectives. (Tania Israel, 2023) Greater Good Science Center at UC Berkeley
• What Having a Growth Mindset Actually Means. (HBR, 2016) Harvard Business Review

this year’s supercomputing conference (sc24) in atlanta was a whirlwind! with 17,000 attendees, it was a massive gathering of the high-performance computing (hpc) community – a community i’ve been a part of since 2013. this year felt particularly special, not just because of the sheer scale of the event, but also because of the significant shift towards ai and the prominent role my team and i played.

The Gnosis Research Center Takes on SC24

sc24 was a major event for the gnosis research center, which i co-founded and currently direct. i was joined by five of my students and a fellow faculty member, and we were thrilled to have a strong presence at the conference. we presented papers, held posters, led workshops, participated in panels, and had a fantastic time networking with colleagues and peers.

two papers anchored part of that program: megammap, which explores the boundary between memory and storage for data-intensive workloads, and dftracer, which makes workflow data movement easier to trace and analyze. the publication records carry the authorship, abstracts, artifacts, and durable citation details; what mattered to me in atlanta was watching those ideas become conversations with the community.

i personally had the honor of co-chairing the pdsw workshop on parallel data storage, a topic that’s incredibly important in the era of ai and big data. i also presented our iowarp project at the hdf5 user group meeting, showcasing some of the exciting research we’re doing at gnosis.

seeing my students actively participate and contribute to the conference was incredibly rewarding. we even had a poster nominated for the best poster award! it’s through events like sc24 that we can foster the next generation of researchers and innovators in hpc.

The Rise of AI in HPC

one of the most striking trends at sc24 was the undeniable convergence of ai and hpc. this is a natural evolution, as ai requires the immense computational power that hpc offers, and hpc can benefit greatly from the intelligence ai brings to scientific workflows.

this shift towards ai is a game-changer. for the first time, we have tools that can generate new content, act as intelligent assistants, and automate tedious tasks. this opens up incredible possibilities for scientific discovery and exploration.

imagine, for instance, using ai in climate science. we can feed ai massive datasets from sensors, satellites, and simulations. the ai can then process this information, generate insights, and even formulate hypotheses for the next iteration of climate modeling. this is just one example of how ai is becoming deeply ingrained in scientific workflows, both for learning and inference.

the growing presence of major players like nvidia, google, and microsoft at sc24 further underscores this trend. these companies are pushing the boundaries of ai and hpc, and their engagement with academic institutions and research labs is fueling innovation in the field.

Supercomputing Goes Mainstream

another key takeaway from sc24 was the sense that hpc is finally breaking out of its niche and entering the mainstream. with the rise of ai, the need for supercomputing resources is becoming more widespread, and this is attracting attention from a broader audience.

this is a positive development for several reasons. first, it exposes more people to the exciting world of hpc, potentially attracting new talent to the field. we desperately need more skilled researchers and engineers in hpc, and this increased visibility can help address the talent gap.

second, a wider audience means more scrutiny and diverse perspectives, which can lead to innovative solutions and accelerate progress. as hpc becomes more critical to various industries and applications, the collective effort to overcome challenges and push the boundaries of what’s possible will only intensify.

Tips and Hints for Conference Goers

attending a conference like sc24 can be overwhelming, especially for newcomers. here are a few tips to help you make the most of your experience:

• prioritize networking: focus on events where you can interact with people, such as workshops, panels, and social gatherings. don’t get bogged down in technical talks – those are usually available in the proceedings afterwards. coffee breaks are a great opportunity to strike up conversations and meet new people.
• pace yourself: sc is a marathon, not a sprint. get plenty of rest, stay hydrated, and bring healthy snacks to keep your energy levels up.
• be curious: don’t be afraid to approach people and ask about their work. most people are passionate about their research and happy to share their insights.
• come prepared: bring business cards to exchange contact information quickly and easily.
• promote your work: if you’re presenting at the conference, use social media to generate interest and draw people to your session.

Looking Ahead

sc24 was a fantastic experience, and i left feeling incredibly energized about the future of hpc and ai. there are still challenges to overcome, particularly in the area of data management for ai workloads. however, the energy and innovation on display at sc24 give me confidence that we’re on the right track.

i’m already looking forward to sc25!

What DLIO does

Deep-learning training does not stress storage in one universal way. Datasets, worker counts, epochs, checkpoints, metadata operations, and framework-specific access patterns all change the picture. Running an entire training stack to study each storage design is expensive and hard to reproduce.

DLIO makes that behavior configurable. You describe the pattern you want to exercise and it generates that pattern, repeatably, without the model. The CCGrid 2021 paper explains the design and the experiments behind it; it won the best paper award, which was a good day.

Why the MLPerf integration matters to me

A benchmark becomes real when a second community uses it to compare systems they built. MLPerf Storage has to exercise storage under defined workload configurations, and its public implementation names DLIO as the generator for those paths. Anyone can inspect the dependency, the configurations, and the rules.

Three things I take from this. DLIO left the environment that produced the paper; other people install it, configure it, and change it without our private context. Integration sharpened the interfaces, because assumptions that felt harmless in one lab became bugs across submitters’ platforms. And the benchmark now sits in a longer evidence chain: MLPerf defines the process, DLIO produces the workload, a submitter supplies the system.

Credit

DLIO started as collaborative research by Hariharan Devarajan, Huihuo Zheng, me, Xian-He Sun, and Venkatram Vishwanath. My part was shaping the benchmark and the systems argument around it. The Argonne team, contributors, and the MLCommons Storage community did the less visible work of turning a research benchmark into shared infrastructure. Being part of MLPerf does not make DLIO a standard by declaration; it means the code was useful enough for another community to build on. That is the outcome I want from research software.

What we are building

IOWarp bridges HPC, AI, and data analytics with one idea: scientific workflows need a context layer, not only a faster byte pipe. The runtime moves data across storage tiers, reshapes it across formats, and exposes it to agents and people through tools with explicit contracts. It builds directly on Hermes, our multi-tier buffering system, and on years of work on labels, tracing, and task-based I/O.

The original announcement mentioned a specialized language model for natural-language data exploration. The plan evolved. What shipped instead is the CLIO ecosystem: a context store, a tool kit of MCP servers, a science agent, and a coding agent, which you can read about in the CLIO story.

What the award changes

Five years is enough time to do systems work properly: architecture, interfaces, implementation, integration, evaluation, documentation, releases, and time for people outside the lab to find the assumptions we missed. It also means students can own substantial pieces of the platform rather than isolated experiments.

An award is runway. The proof is in releases, and that is what the IOWarp organization has been accumulating since.

Updated August 2026 with links to what the project became.