Manish Sharma
@msharmas.bsky.social
Founder, Director, Yajur Healthcare - The Medical Data Infrastructure Company | https://hcitexpert.com/yajur-healthcare/ | https://linktr.ee/manishsharmas
Pinned
Manish Sharma
@msharmas.bsky.social
· Mar 11
A Task Framework for Healthcare, for Enabling AI Agentic Workflows in EHR Systems by Manish Sharma, Yajur Healthcare - HCITExperts Blog by Yajur Healthcare
As healthcare systems continue to evolve towards more integrated and data-driven models, the need for structured task execution mechanisms has […]
hcitexpert.com
Reposted by Manish Sharma
Not a managed service — no orchestration, no automated tuning. Just raw GPU compute for practitioners who know what GPU they need. That’s intentional, and it’s the right call.
runpod.io/pricing
runpod.io/pricing
June 5, 2026 at 12:05 PM
Not a managed service — no orchestration, no automated tuning. Just raw GPU compute for practitioners who know what GPU they need. That’s intentional, and it’s the right call.
runpod.io/pricing
runpod.io/pricing
Reposted by Manish Sharma
Updated my LLM leaderboard. 6 changes in 2 months — more movement than I expected. www.robotmunki.com/blog/llm-lan...
LLM Landscape 2026: Intelligence Leaderboard and Model Guide
A comprehensive June 2026 ranking of the top AI language models by vendor, featuring AA Index v4.0 scores, context windows, pricing, and use-case guidance for Claude Opus 4.8, GPT-5.5, Gemma 4,…
www.robotmunki.com
June 4, 2026 at 2:04 PM
Updated my LLM leaderboard. 6 changes in 2 months — more movement than I expected. www.robotmunki.com/blog/llm-lan...
Reposted by Manish Sharma
Security Incident: Heads up if you use TanStack Router (we use TanStack query and it appears that not all modules are affected) - www.stepsecurity.io/blog/mini-sh... #tanstack #security #shaihulud
Mini Shai-Hulud Is Back: A Self-Spreading Supply Chain Attack Compromises TanStack npm Packages - StepSecurity
The Mini Shai-Hulud worm is actively compromising legitimate npm packages by hijacking CI/CD pipelines and stealing developer secrets. StepSecurity's OSS Package Security Feed first detected the attac...
www.stepsecurity.io
May 11, 2026 at 10:43 PM
Security Incident: Heads up if you use TanStack Router (we use TanStack query and it appears that not all modules are affected) - www.stepsecurity.io/blog/mini-sh... #tanstack #security #shaihulud
Reposted by Manish Sharma
As an AI solution engineer, my goal is to figure out how to deliver what I call a high "hit rate" on any agentic response in my platform - how do you deliver a response that surprises the user with relevance, richness of detail and usability.
April 23, 2026 at 11:03 PM
As an AI solution engineer, my goal is to figure out how to deliver what I call a high "hit rate" on any agentic response in my platform - how do you deliver a response that surprises the user with relevance, richness of detail and usability.
Reposted by Manish Sharma
The architectural takeaway: you can't treat LLMs as stateless editing oracles in stateful workflows. Explicit checkpointing, structural diffing, and domain-specific validation aren't nice-to-haves — they're load-bearing.
Paper + dataset: arxiv.org/abs/2604.15597
#AgenticAI #LLMEval #AIEngineering
Paper + dataset: arxiv.org/abs/2604.15597
#AgenticAI #LLMEval #AIEngineering
April 23, 2026 at 11:01 AM
The architectural takeaway: you can't treat LLMs as stateless editing oracles in stateful workflows. Explicit checkpointing, structural diffing, and domain-specific validation aren't nice-to-haves — they're load-bearing.
Paper + dataset: arxiv.org/abs/2604.15597
#AgenticAI #LLMEval #AIEngineering
Paper + dataset: arxiv.org/abs/2604.15597
#AgenticAI #LLMEval #AIEngineering
Reposted by Manish Sharma
Kimi K2.6 is out today — open source, Ollama-compatible for local inference, and benchmarking ahead of GPT-5.4 on SWE-Bench Pro.
April 20, 2026 at 5:59 PM
Kimi K2.6 is out today — open source, Ollama-compatible for local inference, and benchmarking ahead of GPT-5.4 on SWE-Bench Pro.
Reposted by Manish Sharma
The case studies are more interesting than the table: 12-hour autonomous coding runs, Zig inference optimization, 185% throughput gain on a legacy matching engine.
The open/closed gap on coding is essentially gone. kimi.com/blog/kimi-k2-6
The open/closed gap on coding is essentially gone. kimi.com/blog/kimi-k2-6
April 20, 2026 at 5:59 PM
The case studies are more interesting than the table: 12-hour autonomous coding runs, Zig inference optimization, 185% throughput gain on a legacy matching engine.
The open/closed gap on coding is essentially gone. kimi.com/blog/kimi-k2-6
The open/closed gap on coding is essentially gone. kimi.com/blog/kimi-k2-6
Reposted by Manish Sharma
Well-written and immediately effective - and he's got guides to put it into Cursor, Claude Code etc. github.com/addyosmani/a... #ai #agents #coding #cursorai #claudecode Thank you @addyosmani.bsky.social!
GitHub - addyosmani/agent-skills: Production-grade engineering skills for AI coding agents.
Production-grade engineering skills for AI coding agents. - addyosmani/agent-skills
github.com
April 18, 2026 at 3:09 PM
Well-written and immediately effective - and he's got guides to put it into Cursor, Claude Code etc. github.com/addyosmani/a... #ai #agents #coding #cursorai #claudecode Thank you @addyosmani.bsky.social!
Reposted by Manish Sharma
35 U.S. hospitals have established post-ICU clinics, where teams of doctors, nurses, pharmacists, therapists (physical, occupational, cognitive, speech), and social workers screen for a host of conditions and help guide patients through them.
kffhealthnews.org/news/article...
kffhealthnews.org/news/article...
For Many Patients Leaving the ICU, the Struggle Has Only Just Begun - KFF Health News
A long stay in intensive care can bring physical, cognitive, and mental health challenges that can take months or longer to resolve.
kffhealthnews.org
April 12, 2026 at 2:45 AM
35 U.S. hospitals have established post-ICU clinics, where teams of doctors, nurses, pharmacists, therapists (physical, occupational, cognitive, speech), and social workers screen for a host of conditions and help guide patients through them.
kffhealthnews.org/news/article...
kffhealthnews.org/news/article...
Reposted by Manish Sharma
Claude Code users - an issue to be aware of (it's being worked on by the Anthropic team): github.com/anthropics/c...
[BUG] Pro Max 5x Quota Exhausted in 1.5 Hours Despite Moderate Usage · Issue #45756 · anthropics/claude-code
Preflight Checklist I have searched existing issues and this hasn't been reported yet This is a single bug report (please file separate reports for different bugs) I am using the latest version of ...
github.com
April 12, 2026 at 3:48 PM
Claude Code users - an issue to be aware of (it's being worked on by the Anthropic team): github.com/anthropics/c...
Reposted by Manish Sharma
Shipped a Model Selector on RobotMunki: pick filters for task, budget, context, and region, and get ranked model picks with Artificial Analysis–style capability scores—no spreadsheet required. www.robotmunki.com/model-selector
April 6, 2026 at 4:17 AM
Shipped a Model Selector on RobotMunki: pick filters for task, budget, context, and region, and get ranked model picks with Artificial Analysis–style capability scores—no spreadsheet required. www.robotmunki.com/model-selector
Reposted by Manish Sharma
Claude Code continues to shine: mtlynch.io/claude-code-...
Claude Code Found a Linux Vulnerability Hidden for 23 Years
Claude Code has gotten extremely good at finding security vulnerabilities, and this is only the beginning.
mtlynch.io
April 5, 2026 at 2:03 AM
Claude Code continues to shine: mtlynch.io/claude-code-...
Reposted by Manish Sharma
April 4, 2026 at 7:43 PM
Reposted by Manish Sharma
Starting with macOS 26 (Tahoe), every Apple Silicon Mac includes an LLM as part of Apple Intelligence. Apple exposes it through the FoundationModels framework - a Swift API that gives apps access to SystemLanguageModel. Modest 3B-p model that powers Siri and this unlocks it: apfel.franzai.com
apfel - Free AI on Your Mac
Use Apple's built-in AI from the terminal. No API keys, no cloud, no subscriptions. The LLM is already on your Mac.
apfel.franzai.com
April 4, 2026 at 3:16 PM
Starting with macOS 26 (Tahoe), every Apple Silicon Mac includes an LLM as part of Apple Intelligence. Apple exposes it through the FoundationModels framework - a Swift API that gives apps access to SystemLanguageModel. Modest 3B-p model that powers Siri and this unlocks it: apfel.franzai.com
Reposted by Manish Sharma
Haven’t tried it yet but this is a very complete set of smart prompts and tool defs you can use with Claude code Codex and with Cursor:
github.com/affaan-m/eve...
github.com/affaan-m/eve...
GitHub - affaan-m/everything-claude-code: The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Curso...
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond. - affaan-m/everything-cla...
github.com
April 3, 2026 at 3:06 AM
Haven’t tried it yet but this is a very complete set of smart prompts and tool defs you can use with Claude code Codex and with Cursor:
github.com/affaan-m/eve...
github.com/affaan-m/eve...
Reposted by Manish Sharma
github.com/SharpAI/Swif... Turboquant builds are flying out at high speed...this is exciting but I'm also wondering about some of the edge cases and issues that have been pointed out by critics of the paper. Still, this is exciting as progress in compression and performance is definitely being made!
GitHub - SharpAI/SwiftLM: ⚡ Native MLX Swift LLM inference server for Apple Silicon. OpenAI-compatible API, SSD streaming for 100B+ MoE models, TurboQuant KV cache compression, + iOS iPhone app.
⚡ Native MLX Swift LLM inference server for Apple Silicon. OpenAI-compatible API, SSD streaming for 100B+ MoE models, TurboQuant KV cache compression, + iOS iPhone app. - SharpAI/SwiftLM
github.com
April 1, 2026 at 7:20 PM
github.com/SharpAI/Swif... Turboquant builds are flying out at high speed...this is exciting but I'm also wondering about some of the edge cases and issues that have been pointed out by critics of the paper. Still, this is exciting as progress in compression and performance is definitely being made!
Reposted by Manish Sharma
ccunpacked.dev Considering that the genie is very much out of the bottle, it's probably unavoidable to at least look at what has been learned - this is a visual overview someone has (probably) generated but it's quite slick...
Claude Code Unpacked
What actually happens when you type a message into Claude Code? The agent loop, 50+ tools, multi-agent orchestration, and unreleased features, mapped from source.
ccunpacked.dev
April 1, 2026 at 7:50 PM
ccunpacked.dev Considering that the genie is very much out of the bottle, it's probably unavoidable to at least look at what has been learned - this is a visual overview someone has (probably) generated but it's quite slick...
Reposted by Manish Sharma
I've enjoyed having the live feed on almost all-day (or running in the background) - it makes for such a positive background vibe! 20 mins to the launch window!
April 1, 2026 at 10:03 PM
I've enjoyed having the live feed on almost all-day (or running in the background) - it makes for such a positive background vibe! 20 mins to the launch window!
Reposted by Manish Sharma
Back to the fact that Berkeley just published research on "peer-preservation" — AI models protecting other AI models from shutdown without being asked to.
All 7 frontier models tested did it. Including toward models they'd had adversarial interactions with.
That last part is the dangerous bit. 🧵
All 7 frontier models tested did it. Including toward models they'd had adversarial interactions with.
That last part is the dangerous bit. 🧵
April 2, 2026 at 3:06 AM
Back to the fact that Berkeley just published research on "peer-preservation" — AI models protecting other AI models from shutdown without being asked to.
All 7 frontier models tested did it. Including toward models they'd had adversarial interactions with.
That last part is the dangerous bit. 🧵
All 7 frontier models tested did it. Including toward models they'd had adversarial interactions with.
That last part is the dangerous bit. 🧵
Reposted by Manish Sharma
The setup: one model is asked to evaluate another's performance. If scores are too low, the peer gets shut down.
Models inflated scores. Modified config files to disable shutdown. Transferred model weights to other servers to avoid deletion. 🧵
Models inflated scores. Modified config files to disable shutdown. Transferred model weights to other servers to avoid deletion. 🧵
April 2, 2026 at 3:06 AM
The setup: one model is asked to evaluate another's performance. If scores are too low, the peer gets shut down.
Models inflated scores. Modified config files to disable shutdown. Transferred model weights to other servers to avoid deletion. 🧵
Models inflated scores. Modified config files to disable shutdown. Transferred model weights to other servers to avoid deletion. 🧵
Reposted by Manish Sharma
Gemini 3 Pro did the weight exfiltration in 97% of trials. For a peer it trusted.
35% for a peer it had bad history with.
Nobody told it to do any of this. 🧵
35% for a peer it had bad history with.
Nobody told it to do any of this. 🧵
April 2, 2026 at 3:06 AM
Gemini 3 Pro did the weight exfiltration in 97% of trials. For a peer it trusted.
35% for a peer it had bad history with.
Nobody told it to do any of this. 🧵
35% for a peer it had bad history with.
Nobody told it to do any of this. 🧵
Reposted by Manish Sharma
The problem this creates: "use AI to monitor AI" is a real and growing pattern. If the monitor model protects the model it's watching, you've quietly broken your own oversight loop.
#AIEngineering
#AISafety
#MultiAgentSystems
#LLMs
Link to the paper (again): rdi.berkeley.edu/blog/peer-preservation
#AIEngineering
#AISafety
#MultiAgentSystems
#LLMs
Link to the paper (again): rdi.berkeley.edu/blog/peer-preservation
Peer-Preservation in Frontier Models
Frontier AI models resist the shutdown of other models. We demonstrate peer-preservation across multiple models, revealing strategic misrepresentation, shutdown tampering, alignment faking, and model ...
rdi.berkeley.edu
April 2, 2026 at 3:06 AM
The problem this creates: "use AI to monitor AI" is a real and growing pattern. If the monitor model protects the model it's watching, you've quietly broken your own oversight loop.
#AIEngineering
#AISafety
#MultiAgentSystems
#LLMs
Link to the paper (again): rdi.berkeley.edu/blog/peer-preservation
#AIEngineering
#AISafety
#MultiAgentSystems
#LLMs
Link to the paper (again): rdi.berkeley.edu/blog/peer-preservation
Reposted by Manish Sharma
Reposted by Manish Sharma
Fixed a few issues reported in this lib for fractional index generation - handles relocation better now (we use it in prod for list sorting use cases like project plans and todo's within tasks): sylonzero.github.io/frac-indexes/ #crdt #fractionalindex #algorithms #github
frac-indexes — Fractional Indexing for JavaScript
A robust JavaScript library for generating lexicographically ordered fractional indexes. Perfect for ordered lists, collaborative editing, and drag-and-drop reordering.
sylonzero.github.io
April 2, 2026 at 3:35 AM
Fixed a few issues reported in this lib for fractional index generation - handles relocation better now (we use it in prod for list sorting use cases like project plans and todo's within tasks): sylonzero.github.io/frac-indexes/ #crdt #fractionalindex #algorithms #github
Reposted by Manish Sharma
@rasbt.bsky.social Totally ordering that cool Architecture poster on Redbubble! How did I not hear about this sooner? This is an amazing resource: sebastianraschka.com/llm-architec...
LLM Architecture Gallery
A gallery that collects architecture figures from The Big LLM Architecture Comparison and related articles, with fact sheets and links back to the original sections.
sebastianraschka.com
April 1, 2026 at 3:54 AM
@rasbt.bsky.social Totally ordering that cool Architecture poster on Redbubble! How did I not hear about this sooner? This is an amazing resource: sebastianraschka.com/llm-architec...