@bluesparrow.devAI will not “solve” the mainframe skills shortage. But some expertise can move from tribal knowledge into tools, policies, metadata, runbooks and approval workflows that agents can safely consume. The expert does not disappear. The interface gets better. #Mainframe #MCP
@bluesparrow.devA working MCP connection is not the same as safe automation. Wrong LPAR. Similar job names. Stale output. Partial results. Wrong diagnosis. MCP standardizes tool calls; it does not remove failure modes. Design for predictable constraints, not maximum agent power. #MCP #Mainframe
@bluesparrow.devWould you let an AI coding assistant submit JCL? Don’t start with “AI access: yes/no.” Separate READ, WRITE and EXECUTE. Then decide identity, environment, approvals and audit for each capability. The security boundary is capability by capability. #Mainframe #MCP
@craigmullins.bsky.socialHappy Hug Your Mainframe Day!
October 7th
#mainframe #hugyourmainframe #IBMZ
@bluesparrow.dev“Explain this JCL” is useful. “Why did PAYROLL7 fail this morning?” is more interesting. An agent could gather job status + JES output, correlate it with source, explain the likely cause — then stop for human review. Diagnosis and execution are different capabilities. #Mainframe #MCP
@ithorrorstories.bsky.socialThe software was correct.
The input was correct.
The CPU instruction gave the wrong answer.
IBM 3031, 1978. The investigation went all the way down to single-cycling the processor.
UNPK was broken.
ithorrorstories.eu#ep24
@bibo9.bsky.socialHello everyone,
I found and read this article.
I think it's very interesting because it talks about the truth regarding COBOL, JCL, and DB2. The real problem in the future will be finding someone who knows how to work with them.
#COBOL #mainframe #JCL #DB2
@bluesparrow.devMCP does not replace the mainframe integration layer. The z/OS operation may still be provided through Zowe, z/OSMF REST, RSE API, or another service. The useful part is the boundary: expose narrow agent tools, not generic host access. #MCP #Mainframe
@craigmullins.bsky.socialThis is a nice piece if you are looking for a #mainframe overview
The IBM mainframe: How it runs and why it survives
@bluesparrow.devA coding assistant can explain JCL from source. “Why did this job fail this morning?” is different: it needs runtime context—status, RC, spool, datasets. MCP can expose selected context through controlled tools. The real question: what should the assistant be allowed to retrieve? #Mainframe #MCP
@ithorrorstories.bsky.socialThe software was correct.
The input was correct.
The CPU instruction gave the wrong answer.
Texas, 1978. Robert Garrett single-cycled an IBM 3031 until he found the problem:
UNPK was broken.
Ep24: When UNPK Broke
ithorrorstories.eu#ep24
@raquelsynths.comMISSED_CONNECTION estabelecendo conexão limpa. Deploy de sinal cru, em frequência controlada. Sistemas estáveis em nova transmissão. #mainframe
@derekbrittonuk.bsky.social#mainframe and #opensource have a surprising amount of overlap, but it is not always appreciated. That was the topic of a recent trade show discussion, and the background for this piece written with the super smart Bron Davies at PopUp Mainframe.
@ithorrorstories.bsky.socialMelbourne Cup day, 1991.
Betting system at peak load.
CPU utilization drops to 5%.
Good news ?
Absolutely not.
@triton-consulting.bsky.socialThe GS Forge26 agenda is now live with a packed programme of technical sessions, expert insights & networking with the wider #Db2 community.
We're looking forward to seeing Triton's Rob Gould and Alan Roberts: www.triton.co.uk/gs-forge26-a...
@triton-consulting.bsky.socialAnother excellent London #Db2 User Group #LDUG meeting in the books! Rob Gould and Gareth Copplestone-Jones presented, and Alan Roberts and Iqbal Goralwalla were also representing Triton throughout the day. Thanks to everyone who said hello! See you all at the next one. #IBMZ #Mainframe #IBMChampion