Accendero Software

Full-service operations software support in all major operations areas of interest #businesssoftware #operations #erpsoftware

[Accendero Software @Accendero]

I named orchestration as my bottleneck this week. So how do you get the most for your money?

Where the money goes and the four places to push on it, in this week's newsletter.

#AIAgents #EngineeringLeadership #TechnologyStrategy #ScientificSoftware

https://t.co/sudOdTU99z

[Accendero Software @Accendero]

In this week's newsletter, I discuss how I budget my time and attention, and three places I draw the line.

#AI #EngineeringLeadership #TechnologyStrategy #ChangeManagement

https://t.co/5wVpjuGGmb

[Accendero Software @Accendero]

Five sophisticated ways of reading trial text plus Claude Opus 4.5, and none of them beat a plain bag of words. My limitation was always the data, not the models.

#ClinicalTrials #MachineLearning #AIinPharma

https://t.co/ealfJZ4prU

[Accendero Software @Accendero]

I simulated my own drafting process token by token. The machine's true share came out at 51.5%, a watermark would report about 35%, and Pangram scored this newsletter at 91% AI.

#AI #AIGovernance #AIDetection #ResponsibleAI

https://t.co/Ozoq0Ymtwb

[Accendero Software @Accendero]

Every prior I held about which clinical trial data would predict outcomes turned out to be wrong. The free text beat the registry fields, and the chemistry added almost nothing.

#ClinicalTrials #PredictiveAI #Benchmarking #ClinicalDataScience

https://t.co/NvZILOH8t3

[Accendero Software @Accendero]

Design optimizes toward a target you've already defined. Drug discovery starts by chasing results that don't fit.

#DrugDiscovery #AIinPharma #BiotechInnovation #PharmaRnD

https://t.co/YsBzK4Abnz

[Accendero Software @Accendero]

In this month's Office Hours, we built out a data lakehouse live. This same architecture can transfer and run on AWS at production scale.

#DataEngineering #MachineLearning #ScientificSoftware #DataLakehouse

https://t.co/0Rc7KcX8rq

[Accendero Software @Accendero]

Predicting response from clinical data is one of the hardest problems in healthcare AI.

This is part 1 of a two-part series digging into it. This article covers how I built the benchmark and the topline results.

#ClinicalAI #MachineLearning #DataScience

https://t.co/14B4QzkD4H

[Accendero Software @Accendero]

Everyone's data starts as a mess.

This week's newsletter is the straight line from that mess to being ready to model.

#MachineLearning #DataEngineering #MLOps #DataArchitecture

https://t.co/z2ljedTPQu

[Accendero Software @Accendero]

Anthropic's Claude Science can run a full analysis in minutes. I tested it on the human kinome, a topic from my PhD.

It nailed the execution but answered the wrong question entirely.

#AI #Bioinformatics #AgenticAI #drugdiscovery

https://t.co/MXFKyJ5Vbl

[Accendero Software @Accendero]

For this week's newsletter, I fed Claude six months of my own writing mixed with six months of someone else's on similar topics, then asked it to make a plan to tell us apart and explain how.

#DataScience #MachineLearning #PredictiveModeling #NLP

https://t.co/PJ9odTvvYK

[Accendero Software @Accendero]

Everyone sells the agentic coding loop the same way. Hand it a goal, walk away, come back to something finished.

I spent a week pointing one at a hard biology problem. Maybe science is just different.

#AI #Bioinformatics #AgenticAI #ScientificComputing

https://t.co/C8OrDsteqD

[Accendero Software @Accendero]

Most people I talk with want to use AI but have no idea where to start.

So I wrote The Workflow Automation Playbook for them. A clear, four-phase sequence to get you going.

#AI #GenAI #EnterpriseAI

https://t.co/v75AfP0woE

[Accendero Software @Accendero]

In our June Office Hours session, we walked through the full process of creating a Skill live.

The full recap and a link to the recording are in this week's newsletter.

#ScientificSoftware #AIEngineering #SoftwareStrategy

https://t.co/cdkXIb0u9W

[Accendero Software @Accendero]

An MCP server isn't a config file. It's a live service you operate and babysit forever. Most small teams don't need one yet.

Start with a Skill instead.

#MCP #AIEngineering #ScientificSoftware #AIInfrastructure

https://t.co/p2EM8tJaZZ

[Accendero Software @Accendero]

In this newsletter, we use Claude Opus 4.7 to brainstorm algebra puzzle mechanics, then Claude Code to prototype one.

The result is ugly, the puzzle may not be the right approach, but trying to design it forced me to think about the math.

https://t.co/QKevmHXhil

[Accendero Software @Accendero]

"Buy, don't build."
"Software is essentially free now. Build everything!"

In this week's newsletter, I work through what can be taken from both sides.

#BuildVsBuy #SaaS #SoftwareStrategy

https://t.co/RaMmLi7vxC

[Accendero Software @Accendero]

In this week's newsletter, I walk through how we built a skill that produces validated flow cytometry experimental plans, grounded in CLSI H62.

#ScientificSoftware #LabAutomation #FlowCytometry #AIinScience

https://t.co/ycQ9Is3HXu

[Accendero Software @Accendero]

Maybe I was misinformed about OpenClaw.

I dismissed it at launch. A closer look has changed my read.

#AI #AgenticAI #LifeSciences #Pharma #ScientificSoftware #OpenClaw

https://t.co/sFu3Qz3jSN

[Accendero Software @Accendero]

This week's webinar covered BulkR, a computational method for getting stubborn proteins to crystallize.

#DrugDiscovery #StructuralBiology #ScientificSoftware #ProteinEngineering

https://t.co/LW7egLJ8pX