
Estimated reading time: 3 minutes
THE WORKFLOW
True Signal Media’s FOIA operations routinely involve:
- Reviewing large, multi-agency document productions
- Drafting targeted FOIA requests across federal and state jurisdictions
- Filing administrative appeals citing specific statutory violations
- Tracking 180+ active FOIA requests and 30+ ongoing appeals
- Documenting recurring obstruction patterns across agencies
A single investigation may involve comparing Hawaii’s 21-document SNAP production against Illinois and Colorado responses, identifying withheld record categories, and drafting appeals citing precise statutory obligations.
THE THROTTLING PROBLEM
I maintain paid subscriptions to multiple AI platforms, including Claude (Anthropic) and ChatGPT (OpenAI). Each tool has strengths. Reliability under sustained workload is not one of Claude’s.
Anthropic has implemented usage throttling twice in four months. The operational impact is clear:
- Before throttling: One investigation completed in 1–2 extended sessions
- After throttling: The same work requires 6+ sessions, with lost context between them
Yesterday’s Illinois and Colorado SNAP appeals required:
- Re-establishing prior context from Hawaii’s production
- Re-analyzing disclosed document categories
- Identifying records Illinois and Colorado must possess
- Drafting two appeals with statutory citations
- Adjusting agency-specific procedures
This work was forced into six separate Claude sessions due to conversation limits, each restart increasing error risk and time.
Throttling turns AI from an accelerator into a bottleneck.
CHATGPT’S BULK DOCUMENT ADVANTAGE
When the Minnesota Attorney General produced 68 PDFs related to the Abdifatah Yusuf Medicaid fraud case, the task required systematic analysis to identify withheld materials and build appeal grounds.
ChatGPT workflow:
- Uploaded a single ZIP file containing all 68 documents
- Maintained full conversational context throughout analysis
- Produced categorized findings, gap analysis, and exemption challenges
The result was a comprehensive administrative appeal citing specific document categories and improper withholding. No interruptions. No lost context.
Claude cannot currently process ZIP files or sustain comparable bulk analysis without session resets.
THE VALUE GAP
Anthropic offers:
- $20/month tier with throttling
- $100/month professional tier
The issue is not pricing. It’s functionality.
The higher tier increases message limits and priority access but does not address:
- Bulk document processing
- Context persistence across usage limits
- Workflow continuity for document-heavy investigations
For investigative operations, paying five times more for incremental message increases does not solve the underlying problem.
WHY THIS MATTERS
These are not convenience issues. They are capacity constraints.
When agencies employ obstruction tactics like request splitting, tracking number manipulation, or blanket “no records” claims, investigators need tools that maintain context across prolonged, multi-agency battles.
Context loss mid-investigation leads to:
- Re-explaining obstruction patterns
- Increased risk of citation errors
- Longer appeal timelines
- Fewer total appeals filed
ChatGPT enabled a full analysis of 68 Minnesota documents in a single session. Claude required six sessions for materially simpler comparative work.
RESOURCE LIST DECISIONS
True Signal Media maintains an Allies & Resources page recognizing tools that materially support accountability journalism.
Anthropic is not listed. This is not a judgment of capability, but of reliability. Tools that interrupt active investigations do not qualify as operational allies.
CONCLUSION
AI platforms are now part of investigative infrastructure. Reliability matters more than theoretical capability.
When one tool sustains full-context analysis across large document sets and another fragments work into six sessions, the difference is measurable: fewer appeals filed, fewer agencies challenged, fewer obstruction patterns documented.
In accountability journalism, reliability beats potential every time.