Steven (Batman) Batchelor-Manning
@S_BatMan

Agent Shepherd | Founder | CTO | Building Exceptional Technology

 Bratislava, Slovakia
sbatman.com
Joined September 2010
375 Photos and videos
Best bit about AI, I'm being productive even when I'm on the shitter 🚽
5
1
14
490
When your collection of interlinked notes has grown so vast you can almost start building a galaxy. Adding useless but pretty visualizations to things is the best part of any project
81
Steven (Batman) Batchelor-Manning retweeted
Charles Curran
@charliebcurran
Sep 29
I used AI to explain the AI Doomer drama, with Lord of the Rings and kittens.
351
975
6,196
533,496
This particular agent has supplemented and boosted our teams output drastically since it came online in may. delivering multiple features to the platform, managing ops and leading review etc. Scary how much has changed.
1
5
268
more replies
You research has been inspiring as I am trying to build up my own flow of llm wiki but a with RAG and graph support. It very interesting to play around with.
1
1
48
Glad to hear that it's had a positive impact. I look forward to hearing about your adventures if you tweet about them
4
Looking at solutions to make your agents smarter? Getting the memory system right and getting the data to the agent at the right time is critical. There are a few ways of achieving this.
Steven (Batman) Batchelor-Manning
@S_BatMan
May 21
Article

Almost every serious memory system made the same retrieval decision. Here's why

Ask a flat-vector retrieval system to find the note that mentions the string `idx_memory_units_text_search`. The embedding model has no privileged representation for an arbitrary identifier. The

4
14
571
more replies
A little birdie told me you didn’t work on the shitter, so thought I give you something to fill that slot……..
1
1
25
LMAO, thanks :)
16
Replying to @S_BatMan
Simon Mallas
@thelaserdoctor
Sep 25
Article

The Quiet Third School: Queryable Envelope Memory for Small Agent Teams

The AI-memory debate has two loud schools and, between them, a third one nobody is selling — because you can't put it in a pricing tier. Written from a small team that runs it daily. Honest about

1
1
27
I disagree on the "two loud schools" i'd say there are a few more, but i'll defo give this a read thanks for sharing
1
1
23
Replying to @S_BatMan
Memory quality depends more on retrieval than storage. Semantic search helps, but agents also need exact lookup for IDs and entities, recency filters, and access to structured state
1
1
17
Provenance, agentic memory tools, superseding etc etc all build towards a reliable system that you can trust.
1
9
Different profiles allows you to quickly lean on the best model for the job across subscriptions / keys / local hardware. Automatic fallback, keeps the conversation going during outage or limits being reached
7
311
You only get so much value out of that 1M context, and its a lot less than you think
Steven (Batman) Batchelor-Manning
@S_BatMan
Jun 28
Article

You bought a 1M context. You got 50x less than you paid for.

Every vendor's headline context number is a lie. Not a small lie. A 50 to 100 times lie, depending on what the model is being asked to do. The architecture accepts the input. The model does not read

1
5
443
Are you using the right harness ?
Steven (Batman) Batchelor-Manning
@S_BatMan
Jul 4
Article

You don't have a context problem. You have a harness problem

The harness effect on real workloads runs 5% to 40%. The model effect between frontier models on the same benchmark runs 5% to 15%. The scaffold matters more than what's inside it. The cleanest single
1
139
Is it actually possible to run an entire company with AI agents replacing humans? I mean truly replacing them. Customer service, marketing, sales, operations, etc.
74
29
3,640
We already substantially supplement in commercial, in technical, and everything in between. Almost every member of our company has a dedicated assistant or assistants that are backed by memory-aware AI agents. Most staff are lifting 5x or more above what they would without them. We absolutely haul ass for a medium-sized business
23
Replying to @lamininx
An unreleased locally running piece of software runs from Docker, as a desktop app on three main platforms or as a service on Linux. Specialist harness built off of extensive agentic memory research
1
32
It's a downtime project around my normal business activities but it's helping boost my other businesses substantially
1
17
A junior developer accidentally deployed to production. The homepage is broken. You have 10 minutes. What do you do?
118
31
8,453
Leave them to it, no point getting in the way of learning
3
237
Not only code but the commit messages are getting written by AI these days 🫠
Harry Ashton
@TheHashton
Sep 23
What percentage of your codebase is now written by AI? Mine is 90%, but the number keeps rising.
1
115
more replies
100% agree! Especially when you need to go back months later and figure out why a change was made. I usually ask the agent to summarize the changes and write a semantic commit message. Cursor now even has a magic button next to the commit message box.
1
1
16
When working agentically it's usually a "lint, commit and push" and with a decent git markdown file in the project docs it understands what's expected.
1
6
Steven (Batman) Batchelor-Manning retweeted
Matt Pocock
@mattpocockuk
Sep 23
Stop caring so much about model releases. Focus on the harness, and improving the environment your agent operates in. You'll find yourself far ahead of the curve.
142
147
3,962
168,361
Article

You don't have a context problem. You have a harness problem

The harness effect on real workloads runs 5% to 40%. The model effect between frontier models on the same benchmark runs 5% to 15%. The scaffold matters more than what's inside it. The cleanest single

7
20
189
125,945
Nice work man 👌🏻
1
1
17
cheers :)
19
Load more